absolute frequency | The number of cases or respondents appearing in each category of a frequency distribution or an each cell of a cross-tabulation table.
|
|
|
|
absolute value | The value of a number, ignoring
the plus or minus sign.
|
|
|
|
accessibility bias | One type of selection bias during sampling, where some respondents in the
population are over- or underrepresented because they are more accessible or are less accessible than others.
|
|
|
|
acquiescence bias | Agreement or consent by
respondents to what they believe an interviewer
or sponsor would like them to think or say rather
than giving their real opinion or reaction.
|
|
|
|
action component | One of the three basic components
of an attitude, indicating the person’s
consistent tendency to take action regarding the
topic or to remain passive about it.
|
|
|
|
adjective checklist | A scaling device that lists
a series of adjectives that might be used to describe
some person, place, or thing and asks the respondent
to check any adjective that applies.
|
|
|
|
affiliations | The network of durable, formal, and
informal associations an individual has with family,
relatives, friends, and acquaintances.
|
|
|
|
affinity bias | A form of interviewer bias resulting
from interviewers showing preference for certain
types of people for whom they have an affinity,
such as respondents who are similar to them or
that they find attractive, and including them in
the sample at higher rates than others.
|
|
|
|
aided recall | A form of questioning respondents
about what they remember, where their memory
is aided by presenting or describing the things they
might recall.
|
|
|
|
all-inclusive categories | Response categories
defined to ensure that every feasible answer will
fit into a category; that there will be a category for
every possible answer.
|
|
|
|
alpha level | The critical value, or probability level
above which a relationship between variables will
not be regarded as statistically significant because it is too likely that it could result only by chance from
sampling error if the variables were actually not related
in the population as a whole; also refers to the
probability of a Type I error in academic research.
|
|
|
|
alphanumeric variable | A variable whose
"values" are characters (words or letters) rather
than numeric values.
|
|
|
|
alternative hypothesis | The proposition that
some condition or relationship exists, accepted in
scientific or academic research, if the results fail to
support the "null hypothesis" that it does not exist.
|
|
|
|
analysis of variance (ANOVA) | A statistical
measure of the association between a categorical
independent variable and a continuous, numerical,
dependent variable from an interval or ratio scale,
used to assess the significance of differences among
means for different groups.
|
|
|
|
area sampling | A form of cluster sampling where
the region of the population is first divided into
areas, some are randomly selected, and then
respondents within those areas are randomly
selected.
|
|
|
|
attitude | Relatively durable, psychological predispositions
of people to respond toward or against an
object, person, place, idea, or symbol, consisting of
three components: their knowledge or beliefs, their
feelings or evaluations, and their tendency toward
action or passivity.
|
|
|
|
attitude scale | A scale used to measure attitudes,
usually focusing on the respondents’ feelings or
evaluations toward one or more objects or copies.
|
|
|
|
auspices bias | The tendency for respondents
to react toward the survey sponsor, if they know
who the sponsor is, rather than providing their
own, honest reactions to the survey questions
themselves.
|
|
|
|
average | A measure of central tendency that represents
the most typical case, usually referring to the
arithmetic mean, but also applying to the median
and the mode.
|
|
|
|
banner | A method for showing several
cross-tabulations in one, condensed table in
order to save space or facilitate comparison,
ordinarily used only when one variable is
cross-tabulated against several others.
|
|
|
|
bar chart | A graphic portrayal of several quantities,
such as frequencies or percentages, where the
length of the horizontal or vertical bars represents
the relative magnitudes of the values.
|
|
|
|
behavior | The actions of people or objects in the
past or at the present.
|
|
|
|
bias | The tendency for some extraneous factor
to affect the answers to survey questions or the survey
results in general, in a systematic way, so that
results are "pushed" or "pulled" in some specific
direction.
|
|
|
|
bimodality | The existence of two modes or peaks
in a distribution of response, rather than a single
modal value, often caused when the sample contains
two, distinct populations or groups with
differing reactions.
|
|
|
|
bipolar adjectives | A pair of adjectives, such as
those used in a semantic differential scale, that
represent the polar extremes on one dimension
or continuum.
|
|
|
|
bivariate relationship | A relationship between
only two variables.
|
|
|
|
bivariate statistics | The statistics used to measure
the relationship between only two variables and
assess its statistical significance.
|
|
|
|
bottle scale | A pictorial scale showing a series
of bottles filled to varying levels, sometimes used
when surveying those who may not always understand
verbal or numeric scales, such as young
children or those with reading impairments.
|
|
|
|
breakdowns | A data analysis procedure that
computes and reports the mean, standard deviation,
and number of cases for a continuous numeric
variable for each level of a categorical variable,
ordinarily using analysis of variants to measure
the statistical significance of differences in mean
values.
|
|
|
|
callbacks | The second and subsequent attempts
to contact respondents by telephone or in person
when they were not present to respond to the first
attempt to contact them.
|
|
|
|
case | A set of data obtained from one completed
questionnaire or one respondent that serves as
a single unit for analysis.
|
|
|
|
categorical data | Nominal data where the values
of the variables are merely the names of discrete, independent categories and the numeric magnitudes
have no meaning or stand in no fixed
relationship, as opposed to continuous data.
|
|
|
|
categorical item | A survey question coded with
values that are merely the names of categories, so
that the values do not represent magnitudes or
stand in any ordered relationship with one another.
|
|
|
|
causality | The potential influence or effect that one
item or variable has on another.
|
|
|
|
ceiling effects | The truncation or "chopping off"
of the high side of a distribution because respondents’
answers could go no further up the scale.
|
|
|
|
census | Counting or taking measurements from
all members of a given population, rather than
sampling only a portion to represent the whole.
|
|
|
|
central systems | Computer systems that do the
actual processing at some central location accessible
by communication to several operators or users.
|
|
|
|
central tendency measures | Statistical averages
that describe the most typical value or case, such
as the mean, median, and mode.
|
|
|
|
chi-square | A value, usually obtained from crosstabulation
of two items in survey research, that
can be compared with the values of the chi-square
distribution to obtain a probability for assessing
statistical significance.
|
|
|
|
chi-square distribution | The particular form of
a distribution derived from a set of computations
and defined by the number of "degrees of freedom,"
often listed in statistical reference tables.
|
|
|
|
classification variables | Survey items, such as
demographic variables, that are used to classify
respondents into groups or categories for
comparison.
|
|
|
|
closed-ended questions | A structured survey
question where the alternative answers are listed
so that respondents must ordinarily pick only from
among them.
|
|
|
|
cluster | One group of individuals or sampling
units that have proximity with one another within
the sample frame in some respect, such as those
within a given area.
|
|
|
|
cluster bias | A form of selection bias resulting
when a cluster sampling design selects respondents
who are too closely related to one another within
a cluster, so that they tend to give similar responses.
|
|
|
|
cluster sampling | A technique often used in surveys
to save travel or long-distance toll charges
where the population is divided into clusters and
a few clusters, each containing many respondents,
are randomly selected.
|
|
|
|
codebook | The entire set of codelists for several
variables from one survey.
|
|
|
|
codelist | The list of code values and category
labels for a single survey variable that is generated
by postcoding and used during analysis and
reporting.
|
|
|
|
coding | The process of assigning code values
to the various alternative answers to survey
questions, either when constructing the questionnaire
(precoding) or after the data collection
(postcoding).
|
|
|
|
coefficient of determination | The square of the
correlation coefficient, indicating the proportion
of "shared" variance for correlation or the proportion
of variance in the dependent variable that
was "explained" by the independent variable
in regression.
|
|
|
|
comparative scale | A scale using one entity as the
standard by which one or more others are judged
or evaluated.
|
|
|
|
computer hardware | The physical devices and
components of computers, including both the
central processing unit and the peripheral devices
for inputting, storing, and outputting data.
|
|
|
|
computer software | The programs and coded
instructions to the computer, including both the
operating system that provides general control
and the applications programs that perform specific
computations.
|
|
|
|
conditional branching | Instructions or "go-to"
statements in a questionnaire indicating the interviewer
or respondent should skip items that don’t
apply, based on answers to previous questions.
|
|
|
|
confidence interval | The range around a numeric
statistical value obtained from a sample, within
which the actual, corresponding value for the
population is likely to fall, at a given level of
probability.
|
|
|
|
confidence level | The specific probability of
obtaining some result from a sample if it did not
exist in the population as a whole, at or below
which the relationship will be regarded as statistically
significant.
|
|
|
|
continuous variable | A variable that represents
a continuum without any breaks or interruptions,
so the numeric values could potentially take on
an infinite number of values expressed in whole
numbers and fractions.
|
|
|
|
convenience sample | A sample selected more on
the bias of the researcher or data collection team’s
convenience than on the requirements for random
selection with a known probability of inclusion and
representation.
|
|
|
|
correlation analysis | A measure of the relationship
or association between two continuous numeric
variables that indicates both the direction and degree
to which they co-vary with one another from
case to case, without implying that one is causing
the other.
|
|
|
|
correlation coefficient | The value computed with
correlation analysis, ranging from zero to indicate
no systematic relationship to plus or minus one,
indicating a perfect linear relationship, where the
positive or negative value shows if the relationship
is direct or inverse, respectively.
|
|
|
|
correlation matrix | The correlation coefficients
between each pair, for several variables, arranged
so that each variable is identified on each row and
on each column, with the coefficient listed in the
cells defined by the rows and columns.
|
|
|
|
critical value | The probability level above which a
relationship between variables will not be regarded
as statistically significant because it is too likely that
it could result only by chance from sampling error
if the variables were actually not related in the population
as a whole.
|
|
|
|
cross-tabulation | Plotting two categorical variables
in the form of a matrix so that the values of
one variable define the rows and the values of the
other define the columns, with the cells containing
the frequency of cases with a given value for each
of the two items and from which a chi-square value
can be computed to assess the statistical significance
of the relationship.
|
|
|
|
curvilinear | A line or distribution of values that
is continuous, but forms an arch, rather than a
straight line.
|
|
|
|
data | Most often numbers, but also letters or
words that symbolize or represent quantities,
entities, or categories of things.
|
|
|
|
data analysis | The manipulation of numbers,
letters, or symbols in order to suppress the detail
and reveal the relevant facts or relationships.
|
|
|
|
data collection | The process of communicating
questions and obtaining a record of responses from
a sample, either by mail, telephone, or personal
interviewing.
|
|
|
|
data field | The location and number of columns
in a data file record required to contain the largest
number of digits for any code value for a particular
variable.
|
|
|
|
data point | A datum, or one single entry of a
number, letter, or symbol, usually for one variable
and one case or respondent.
|
|
|
|
data processing | Submitting the survey data to
computer programs and routines in order to perform
the statistical analysis and to generate reports,
as opposed to hand tabulation of the data.
|
|
|
|
decisions | The process and/or results of individual
evaluations, judgments, and choices among
alternatives.
|
|
|
|
degrees of freedom (d.f.) | A parameter most
often based on the number of cases or respondents,
but slightly reduced to adjust for some earlier
computations and used when checking reference
tables or computing probability to assess statistical
significance.
|
|
|
|
demographics | A set of conditions or attributes
of people, often including age, sex, marital status,
education, employment, occupation, and income,
among others, usually measured in surveys to
determine the types of people represented by the
sample and to make comparisons of other results
among demographic groups.
|
|
|
|
dependent variable | The variable that is viewed
as being potentially influenced, affected, or determined
by some other variable in a cause-and-effect
relationship, based on the logic and meaning of the
things represented by the variables.
|
|
|
|
depicted scale | Any scale that cannot be included
within the survey question, so that it must be
shown in the form of numbers, words, or pictures
representing response alternatives.
|
|
|
|
descriptive research | Research that is designed
primarily to describe rather than to explain a set of
conditions, characteristics, or attributes of people in
a population, based on measurement of a sample.
|
|
|
|
descriptive statistics | Statistics such as averages
and measures of spread, used to suppress the
detail in data files and to condense and summarize
the data to make facts more visible, as well as to
indicate the degree to which the sample data are
likely to represent the entire population.
|
|
|
|
desires | Express wishes or conscious urges of
respondents that may be a topic for survey research
measurement within the broader topic category
of "needs."
|
|
|
|
desktop systems | Computer systems that do some
or all of the processing right at the operator’s desk
or work station.
|
|
|
|
diagram scale | Any form of scale that uses
a diagram to depict the response options or
to obtain or collect answers to survey questions.
|
|
|
|
dichotomous question | A question with only two
response alternatives, such as a yes/no question or
an item that can either be checked or ignored.
|
|
|
|
discrete variable | A categorical variable yielding
nominal data, where all of the answers must fall
within a category and the code values stand in
no ordered relationship to one another.
|
|
|
|
discriminant analysis | A statistical measure of
the relationship between a continuous, numeric
independent variable from an interval or ratio
scale and a categorical dependent variable defining
two or more groups, used both to assess statistical
significance and also to compute the discriminant
function, used to predict or classify new cases
into groups.
|
|
|
|
discriminant function | The prediction or classification
equation obtained from discriminant analysis,
used to predict or classify new cases into groups
when only the value of the independent variable
is known.
|
|
|
|
dispersion | The range and degree of spread or
variance in the distribution of data for a survey
variable.
|
|
|
|
editing | The process of examining questionnaires
or data against some set of criteria, to be sure the
content is correct or appropriate.
|
|
|
|
executive summary | The first section of the final
report, including only an outline of the major highlights
of the results.
|
|
|
|
expected cell frequency | A value computed
during or after cross-tabulation, based on the
proportion of the data represented by the entire
row and column on which the cell resides.
|
|
|
|
explicit scale | Any scale that is directly expressed
or stated, either verbally or visually, as opposed to
those that are only implied by the question.
|
|
|
|
extreme case | A response that is an outlier, with
a value so extreme it is far distant from any other
response, which may sometimes suggest there
has been an error while recording or transferring
the data.
|
|
|
|
extremity | The extreme or terminal point, limit, or
part of something, such as the upper or lower limit
of a scale or distribution of scores.
|
|
|
|
F-distribution | The particular form of a distribution
derived from a set of computations and
defined by two numbers of "degrees of freedom,"
often listed in statistical reference tables.
|
|
|
|
F-ratio | The ratio of a numerator and denominator
value consisting of variance expressed as "mean
squares," or the "sum of squares" divided by the
degrees of freedom for each, usually computed
with analysis of variance and compared to the
F-distribution in a statistical reference table to
assess statistical significance.
|
|
|
|
feeling component | One of the three main components
of attitudes, consisting of the evaluations and
judgments of the topic by the individual holding
the attitude.
|
|
|
|
fixed sum scale | A particular type of scale where
the respondent is asked to list the number of times
each of a set of alternatives occurred or apply, out
of a given total, so that the sum of the values must
equal the total.
|
|
|
|
floor effects | The truncation or "chopping off" of
the low side of a distribution because respondents’
answers could go no lower on the scale.
|
|
|
|
forced ranking scale | A type of scale, yielding
ordinal level data, where the respondents
are instructed to rank a series of items in sequential
order, with no "ties" or equal rankings
allowed.
|
|
|
|
formatting | The design or a particular arrangement
of words, numbers, or symbols, specifying
their order or sequence, physical location, relative
distance or proximity, and general form, often used
in reference to the location of data fields within a
file record.
|
|
|
|
frequency distribution | The number of cases that
contain each of the scale values for a particular
survey item or variable.
|
|
|
|
frequency table | A tabular presentation of the
frequency distribution, often including percentage
distributions based on the frequencies and the
sample size, the number of valid cases, and the
cumulative number of valid cases.
|
|
|
|
FTP | Abbreviation of File Transfer Protocol, the
protocol used on the Internet for sending files.
|
|
|
|
goals | Specific objectives or ends sought by
respondents that may be a topic for survey research
measurement within the broader topic category of
"needs."
|
|
|
|
go errors | An error that results when a decision
maker goes ahead with a course of action and it
proves to be costly or unsatisfactory.
|
|
|
|
grand mean | The arithmetic mean of the dependent
variable for all of the cases in an analysis
of variance, as opposed to the "group means,"
including only the cases in each category of the
independent variable.
|
|
|
|
graphic data description | Portrayal of data distributions
for individual survey variables with
charts, graphs, or data plots rather than numeric
tables.
|
|
|
|
graphic scale | Any scale with scale points
portrayed as pictures or diagrams, rather than
numbers, letters, or words.
|
|
|
|
hand tabulation | The statistical computations and
analysis of survey data without the use of computer
analysis routines, usually confined to frequency
tables, cross-tabulation tables, and the statistics
that can be obtained from them without the processing
of continuous numeric variables with many
scale values.
|
|
|
|
happy-sad face scale | One form of pictorial scale
showing a series of simple sketches of faces with
the mouth of each turned up in a smile to show
pleasure or down to show displeasure in varying
degrees, often used for surveying young children
who may not be able to understand numeric or
verbal scales.
|
|
|
|
histogram | A horizontal bar chart showing the
frequency or percentage distribution of response
for a survey item in graphic form, often generated
by computer analysis routines if requested.
|
|
|
|
homoskedasticity | The required condition of
a scatterplot of data for regression analysis, where
the data points are spread around the regression
line in approximately equal amounts at any given
point on the line, forming an even corridor of data,
as opposed to heteroskedasticity, such as a "funnel
shaped" pattern of data round the regression line.
|
|
|
|
hostility bias | Opposition, resentment, or resistance
to an interviewer, survey sponsor, response
task, or situational factor that negatively affects
answers to one or more specific survey questions.
|
|
|
|
HTML | Short for HyperText Markup Language, the
authoring language used to create documents on
the World Wide Web. HTML defines the structure
and layout of a Web document by using a variety
of tags and attributes.
|
|
|
|
hypothesis | A conjectural statement about the
value of some variable or the relationship between
variables that will be tested and ultimately accepted
or rejected on the basis of statistical analysis of
survey results, most often used in formal scientific
or academic research.
|
|
|
|
images | The generalized or synthesized picture
representation of some object, person, or idea held
in the minds of people, based on partial information
from previous experiences, perceptions, or
evaluations, and often one of the major topics of
survey research.
|
|
|
|
implicit scale | Any scale that need not be explicitly
stated in the question or presented verbally or
visually to respondents because they automatically
understand how they are to respond, such as asking
one’s age with the implicit understanding that it
will be expressed in years since birth.
|
|
|
|
independence | The condition between two
variables or measurements where information
about one gives no indication of the likely value
of the other because they are unrelated.
|
|
|
|
independent variable | The variable that is viewed
as influencing, affecting, or determining the values
of another variable when they are regarded as
being in a potential cause-and-effect relationship.
|
|
|
|
inferential statistics | Any statistical measure that
can be used to make inferences or generalizations
about a population, with a known level of probability,
based on the values or conditions of a sample.
|
|
|
|
information needs | The specific categories of
information required by those sponsoring pragmatic
survey research, in order to make decisions
or choices or to set policy, or required by those
conducting academic research to test a theoretical
or conceptual hypothesis and enhance some body
of knowledge or literature.
|
|
|
|
instrumentation | The survey questionnaire and
other devices, such as cover letters, rating cards,
and the like, used to obtain data from respondents.
|
|
|
|
instrumentation bias | The tendency for some
aspect of the survey instruments to cause respondents
to answer in a particular way or systematically
"push" or "pull" the survey results
in some given direction, thus reducing the survey
validity.
|
|
|
|
instrumentation error | The tendency for some
aspect of the survey instruments to randomly affect
the data in such a way that they are not true representations
of the respondent opinions or conditions,
but there is no specific direction or systematic influence,
so that survey reliability is reduced.
|
|
|
|
integer | A whole number, as opposed to a fractional
or decimal value.
|
|
|
|
interpretation error | Error that results when interviewers
are asked to interpret responses during the
interview or make judgments about the responses.
|
|
|
|
interrogation error | Errors that occur when questions
are expressed differently from one respondent
to the next.
|
|
|
|
interval scale | Any scale where the intervals
between scale points are equal, even though there
may be no zero value or zero does not represent
a complete absence of the thing measured, such as
the Fahrenheit scale.
|
|
|
|
interviewing bias | The tendency for some aspect
of the interviewing to cause respondents to answer
in a particular way or systematically "push" or
"pull" the survey results in some given direction,
thus reducing the survey validity.
|
|
|
|
interviewing error | The tendency for some aspect
of the interviewing to randomly affect the data so
they don’t truly represent the respondents’ opinions
or conditions, thus reducing survey reliability.
|
|
|
|
judgment sample | A sample selected on the basis
of the researcher’s judgment about what units or
respondents should and should not be included,
as opposed to random selection.
|
|
|
|
knowledge component | One of the three main
components of attitudes, consisting of the facts
or beliefs the individual holds about the topic of
the attitude.
|
|
|
|
kurtosis | A statistical measure of the shape
of a distribution that indicates whether the
curve is more peaked or more flat than a normal,
bell-shaped curve and how much so.
|
|
|
|
level of confidence | The specific probability of
obtaining some result from a sample if it did not
exist in the population as a whole, at or below
which the relationship will be regarded as statistically
significant.
|
|
|
|
lifestyle | The general pattern of daily behavior, activities,
choices, and preferences for an individual
or family that might be used to characterize them
and distinguish them in meaningful ways from
those following a different pattern.
|
|
|
|
Likert scale | A type of scaling where the respondents
are presented with a series of statements,
rather than questions, and asked to indicate the
degree to which they agree or disagree, usually
on a five-point scale.
|
|
|
|
linear, numeric scale | A scale used when items
are to be judged on a single dimension and arrayed
on a scale with equal intervals, providing both
absolute measures of importance and relative
measures, or rankings, if responses among the
various items are compared.
|
|
|
|
mail data collection | The mailing of questionnaires
and their return by mail by the designated
respondents.
|
|
|
|
maximum | The highest value for a variable
that was actually obtained from a sample, often
reported by analysis routines and used by analysts
to assess range and likelihood of outliers or ceiling
effects.
|
|
|
|
mean | The most common average or measure
of central tendency, providing an indication of the
most typical or representative value for the sample
and the population as a whole, within a given
confidence level.
|
|
|
|
mean squares | A value usually computed for
analysis of variance to form an F-ratio to assess
statistical significance, consisting of the total of the
squared deviations from the mean for each data
point, or sums of squares, divided by the number
of cases or degrees of freedom.
|
|
|
|
measures of spread | Statistical indications of
the dispersion of the data around the central point,
such as the standard deviation.
|
|
|
|
median | An average or measure of central tendency,
consisting of the value the middle case
would take on if the cases were arrayed from
lowest to highest value for the variable and the
scale represented a continuum or could include
an infinite number of points, used in preference
to the mean for ordinal level data and often preferred
to the mean for distributions that are highly
skewed to one side or have outlying values.
|
|
|
|
mental set | The existing frame of mind, point of
view, or train of thought adopted by a respondent
at a given moment, used to judge a series of survey
questions or items.
|
|
|
|
midpoint | The middle point on a scale with an
odd number of scale points, sometimes reflecting
neutrality on the spectrum of response.
|
|
|
|
minimum | The lowest value for a variable
that was actually obtained from a sample, often
reported by analysis routines and used by analysts
to assess range and likelihood of outliers or floor
effects.
|
|
|
|
minimum expected cell frequency | The lowest
expected cell frequency in a cross-tabulation
table, that must be at least five for valid use of
the chi-square statistic to assess the significance
of the relationship, computed by identifying the
smallest row frequency and column frequency,
multiplying the two, and dividing by the total
frequency for the table.
|
|
|
|
mode | The only average appropriate to indicate
the most typical case for a distribution of nominal
data, consisting of the category with the highest
frequency, and also representing the location of the
peak or high point in a distribution of continuous
numeric data with many scale values.
|
|
|
|
motives | The impetus or urge causing a person
to take some action, that may be a topic for survey
research measurement within the broader topic
category of "needs."
|
|
|
|
multiple rating list | A survey item format used
to save space and response time, designed so many
items can be rated on the same scale.
|
|
|
|
multiple rating matrix | A survey item format
used to save space and response time, designed
so the same scale is used to rate multiple items on
several dimensions.
|
|
|
|
multiple regression | Linear regression that uses
a single dependent variable and two or more independent
variables in the same analysis, in contrast
to simple, linear regression using only one independent
variable, so that both the effect of each independent
variable and the effects of interactions
among independent variables can be gauged.
|
|
|
|
multiple response item | A structured, multiplechoice
survey question that allows the respondent
to choose as many response categories as apply.
|
|
|
|
multivariate analysis | Statistical analysis techniques
to assess the relationships or patterns among
more than two variables simultaneously, including
such methods as multiple regression, factorial
analysis of variance, analysis of covariance, factor
analysis, cluster analysis, multidimensional scaling,
and the like.
|
|
|
|
mutually exclusive categories | Response categories
defined to ensure a unique association
between any given answer and only one category
or alternative, so no response can fit into two or
more categories.
|
|
|
|
namelist | A listing of names and addresses, often
used for mail surveys, that may be accumulated
or acquired from one of many namelist brokerage
firms who accumulate and manage such lists for
people with particular characteristics or in certain
locations.
|
|
|
|
nay-sayer | An individual or respondent
who persistently tends to respond in the negative
more often than others, regardless of the
question.
|
|
|
|
nay-sayer bias | The tendency for a set of survey
results to be generally and artificially negative on
a series of items because all items are inclined in
the same direction, toward the positive or toward
the negative, and negative responses to the earlier
items were generalized to the remaining ones, thus
reducing the validity.
|
|
|
|
need | A persistent or fundamental requirement
of the individual in order to maintain physical,
psychological, or social well-being, usually fluctuating
over time in its degree of satisfaction, and
often a topic for survey research measurement.
|
|
|
|
no-go errors | Errors that result when a decision
maker either fails to take some action that would
have positive results, or ignores an alternative
that would be more positive, choosing some less
positive course.
|
|
|
|
nominal scale | A scale that uses numbers, letters,
or symbols only as the names of independent categories,
so that the scale values do not stand in any
ordered relationship to one another.
|
|
|
|
nonprobability sampling | A nonrandom sampling
design such as convenience sampling,
where the probability of selecting a given sampling
unit from the population is neither known
nor equal to the probability of selecting any
other unit.
|
|
|
|
nonrespondents | Those in the population who
were included in the sample but failed to respond
because they refused or could not be reached, or
for some other reason.
|
|
|
|
nonresponse bias | A systematic effect on the
data reducing validity that results when those with
one type of opinion or condition fail to respond to
a survey more often than do others with different
opinions or conditions.
|
|
|
|
normal curve | Any distribution that conforms
exactly or very closely to a normal distribution.
|
|
|
|
normal distribution | A continuous, symmetrical
distribution that forms a curve with a particular
shape defined by a mathematical equation, often
referred to as a "bell-shaped" curve, valuable as a
statistical reference because the precise areas under
the curve can be computed or obtained from statistical
tables.
|
|
|
|
N-size | A commonly used term for the sample size
or the number of cases included in an analysis or
tabular report.
|
|
|
|
nth name sampling | A sampling design where the
number of units in the sample frame is first divided
by the desired sample size to obtain the value of n,
a value between one and n is randomly selected as
a starting point or first case to be selected, and then
every nth name or unit is selected, yielding a random
sample.
|
|
|
|
null hypothesis | The hypothesis stipulating there
will be no significant relationship between two
variables, which can be tested with survey or other
data and rejected in favor of the alternative hypothesis
if the relationship proves significant, and most
often used in scientific or academic research.
|
|
|
|
numeric item | Any survey item with scale numbers
that are meaningful and stand in an ordered
relationship to one another, such as those from
ordinal, interval, or ratio scales.
|
|
|
|
online focus groups | Participants sign in at a
private chatroom on the Internet where they either
join in a focused discussion with a moderator or
read messages on a bulletin board and post their
observations and opinions over a period of several
days or weeks.
|
|
|
|
open-ended question | An unstructured survey
question that does not include a list of alternative
answers, so that respondents must answer in their
own words.
|
|
|
|
order bias | The tendency for the order in which
survey items are listed to affect respondents’ answers
in some systematic way, reducing validity.
|
|
|
|
ordinal scale | A particular type of scale where the
response alternatives define an ordered sequence,
so the first is less than the second, the second less
than the third, and so on, yielding ordinal level data
where the intervals between scale points are not
known or necessarily equal.
|
|
|
|
outlier | An extreme case or data point that stands
well above or well below its nearest neighbor and
is highly atypical of the distribution as a whole.
|
|
|
|
paired comparison scale | A type of scale that
presents respondents with one pair of alternatives
at a time, instructing them to pick just one from
each pair, yielding dichotomous, nominal data.
|
|
|
|
paired t-test | A technique for assessing the statistical
significance of differences in mean values when
both are obtained from the same respondents, and
are therefore paired with one another.
|
|
|
|
panel data collection | A survey of a group of preselected
respondents who agreed to be panel members
on a continuous basis for a given period of
time and provided initial demographic data, allowing
selection of special groups and permitting the
use of surveys to monitor responses over time.
|
|
|
|
parameter | A coefficient or value for the population
that corresponds to a particular statistic from
a sample and is often inferred from the sample.
|
|
|
|
PDF | Short for Portable Document Format, a file
format developed by Adobe Systems. PDF captures
formatting information from a variety of desktop
publishing applications, making it possible to send
formatted documents and have them appear on
the recipient’s monitor or printer as they were intended.
To view a file in PDF format, you need
Adobe Acrobat Reader, a free application distributed
by Adobe Systems.
|
|
|
|
percentage distribution table | A table listing the
percentage of respondents selecting each response
category or scale point.
|
|
|
|
percentile | An indication of the position of a
case or value within a distribution, based on the
number of cases with a lesser value out of a total
of 100 cases.
|
|
|
|
peripheral devices | Units of computer hardware,
other than the central processing unit, used to
input, store, and output data or information.
|
|
|
|
personal interview | Data collection accomplished
with the interviewer in the presence of the respondent,
so that they have visual contact, as opposed
to telephone interviewing.
|
|
|
|
pictorial scale | Any scale with scale points
portrayed as pictures or diagrams rather than
numbers, letters, or words.
|
|
|
|
pie chart | A method for portraying results graphically,
consisting of a circle divided by lines from the
center to the perimeter, so that the angles between
the lines and therefore the size of the "pieces" represent
proportions.
|
|
|
|
pilot survey | A brief preliminary survey, often
using a small, convenience sample, conducted
to test the survey instruments and data collection
method before the project details are finalized and
the larger, formal survey conducted.
|
|
|
|
population | The definition of all those people or
elements of interest to the information seekers and
from among whom the sample will be selected.
|
|
|
|
population parameters | Values or coefficients such
as the mean or variance that describe the distribution
of a variable in the population, often estimated
or inferred based on the corresponding values of
sample statistics.
|
|
|
|
postcoding | The process of examining completed
survey questionnaires, choosing response categories
for items not precoded, assigning code
values to them on the documents, and recording
codes and category labels in a codelist.
|
|
|
|
precision | The range of the confidence interval
at a given level of probability, expressed in absolute
terms or as a percentage of the mean value.
|
|
|
|
precode | Assigning code values to the categories
of structured questions and listing them for printing
on the questionnaire prior to data collection.
|
|
|
|
preferences | Predetermined choices by respondents
from among alternative goods, that may be
a topic for survey research measurement within
the broader topic category of "needs."
|
|
|
|
prestige | A condition of superior status, rank, or
distinction relative to one’s peers or society in general,
constituting a basic human need, sometimes
causing respondents to react to questions in ways
they perceive to be more prestigious.
|
|
|
|
pretest | Preliminary trial of some or all aspects
of the sampling design, survey instrumentation,
and data collection method, to be sure there are
no unanticipated difficulties or problems.
|
|
|
|
primary data | Data collected for a particular project
to meet specific information needs, as opposed
to data that already exist for general use or as the
result of inquiries for other purposes.
|
|
|
|
probability sampling | Any sampling design
where every element in the population has either
an equal probability of selection, as with random
sampling, or a given probability of being selected
that is known in advance and used in analysis to
assess significance.
|
|
|
|
process editing | Examining survey data with
computer processing routines to be sure the data
conform to the data file format and that all values
are expressed in the proper form and are within
the range of the scale for each item.
|
|
|
|
product-moment correlation | The statistical
method of correlation that requires interval or ratio
level data and is not appropriate for ordinal scale
data, which require rank correlation.
|
|
|
|
qualification | The process of inspecting or interrogating
potential respondents to be sure they
are qualified to respond or that they fit the quota
specifications for a particular interviewer.
|
|
|
|
qualitative research | Research obtaining data in
the form of words or other indications that do not
lend themselves to quantitative analysis and whose
analysis and interpretation depend on subjective
judgments by experts.
|
|
|
|
quantitative research | Research obtaining data
in a form that can be represented by numbers, so
that quantities and magnitudes can be measured,
assessed, and interpreted with the use of mathematical
or statistical manipulation.
|
|
|
|
questionnaire | The basic survey instrument containing
instructions, questions, or items, response
alternatives where appropriate, and specific means
for recording responses.
|
|
|
|
quota | A set number or proportion of respondents
with given characteristics or attributes sought in
a sample or assigned to specific interviewers or
field-workers.
|
|
|
|
quota sample | Any sampling design that requires
a set number or proportion of respondents with
given characteristics or attributes.
|
|
|
|
quota specification | The listing of quota requirements
for the entire sample or for specific
interviewers, including identification of the
characteristics that define the quota, the manner
in which they are to be ascertained, the method
of qualification of respondents, and the number
or proportion of respondents who are to have each
attribute or combination of attributes.
|
|
|
|
random digit dialing | A sampling system for
telephone surveys where all telephone numbers
in households or all that have one of a given set of
three-digit telephone number prefixes are regarded
as the sample frame, and seven-digit or four-digit
numbers are generated and dialed manually or
automatically to obtain the sample.
|
|
|
|
random error | The result of extraneous factors,
such as sampling error, affecting the survey
results in no systematic pattern, so the answers
are not consistently pushed or pulled in one
specific direction.
|
|
|
|
random sampling | A sampling design that
seeks to select respondents from the population
or sample frame in a completely random fashion,
so every respondent has an equal probability of
being selected.
|
|
|
|
range | A measure of the spread in the distribution
of data for a variable, defined as the maximum
minus the minimum, plus one.
|
|
|
|
rank correlation | The statistical method of correlation
appropriate when one or both of the variables
are from only ordinal level scales, sometimes called
Spearman rank correlation.
|
|
|
|
rank order scale | A scale essentially the same as
the forced ranking scale.
|
|
|
|
rating card | A card or sheet containing a rating
scale that is handed to or shown to respondents
during personal interviews and from which they
pick their response alternatives by number or letter.
|
|
|
|
rating scale | Any scale from which respondents
choose values that represent their responses, ordinarily
yielding interval or ratio level data.
|
|
|
|
ratio scale | Any scale that has the same characteristics
as an equal interval scale, plus the fact
that zero represents the complete absence of the
thing being measured, so that a ratio of one scale
value to another has meaningful and legitimate
interpretation.
|
|
|
|
raw data | Data that has not been transformed or
processed, although it may have been edited and
transferred from one medium to another.
|
|
|
|
recode | The process of systematically assigning
new code values to variables, based on the original
values, usually done in order to group data into
larger categories to obtain fewer code values.
|
|
|
|
record format | The specification of where the data
field for each variable is to be keyed or recorded in
a data file, including both the column(s) and the
record numbers within a single case.
|
|
|
|
recording error | Error that may occur when interviewers
are required to write down verbal answers
by respondents, typically caused by unsatisfactory
abbreviation when there is insufficient time to
record entire verbatim answers.
|
|
|
|
regression analysis (simple linear regression) | A statistical measure of the effect of one interval or
ratio level variable on another, used both to indicate
the statistical significance of the relationship and
to generate an equation to predict or estimate the
value of the dependent variable for a new case,
based only on the known value of the independent
variable.
|
|
|
|
regression equation | The equation generated by
linear regression analysis, expressed as a coefficient
that can be multiplied by the value of the independent
variable for a new case and a constant to be
added to predict the unknown value of the dependent
variable.
|
|
|
|
relative frequency | A term that is sometimes used
to refer to the percentages listed in a frequency
table, indicating the proportion of the sample in
each category.
|
|
|
|
reliability | The degree to which the survey results
are free from random error, as opposed to systematic
bias, often expressed in terms of confidence
intervals or confidence levels.
|
|
|
|
report generation | The process of arranging and
condensing tabular survey results and expressing
the written interpretations of the findings to provide
information to those seeking it.
|
|
|
|
responding sample | The number of cases
with valid responses to the survey or to an individual
survey item, as opposed to the total sample
size.
|
|
|
|
response bias | The tendency for some aspect of
the response task, such as annoyance or a desire
to please the interviewer, to cause respondents to
answer in a particular way or systematically
"push" or "pull" the survey results in some given
direction, thus reducing the survey validity.
|
|
|
|
response error | The tendency for some aspect of
the response task, such as boredom, inattention, or
fatigue, to randomly affect the data in such a way
that they are not true representations of the respondent
opinions or conditions, but there is no specific
direction or systematic influence, so that survey
reliability is reduced.
|
|
|
|
response option error | Error resulting because
interviewers read the alternative answers to respondents
when they shouldn’t do so or fail to read
them when they should.
|
|
|
|
response rate | The percentage of those included
in the sample who responded to the survey and
provided usable, completed questionnaires.
|
|
|
|
r-square (r2, Rsq) | The coefficient of determination
obtained during regression analysis, indicating the
proportion of variance in the dependent variable
that is "explained" by the values of the independent
variable.
|
|
|
|
runs test | A statistical process used in connection
with regression analysis to determine the probability
that the data are actually linear or arrayed
evenly around a straight line if the data were plotted,
by counting the "runs" of successive data
points that are all on one side of the regression line.
|
|
|
|
sample | The number and/or identification of
respondents in the population who will be or have
been included in the survey.
|
|
|
|
sample frame | A listing that should include all
those in the population to be sampled and exclude
all those who are not in the population.
|
|
|
|
sample selection bias | Any form of bias resulting
from the selection of respondents in a manner that
deviates from random selection, so that some types
of respondents are over- or underrepresented in the
sample.
|
|
|
|
sample unit | The smallest unit of the sample
to be surveyed or the unit that will constitute one
case for analysis, ordinarily one respondent or
questionnaire.
|
|
|
|
sampling design | The specification of the sample
frame, sample size, and the system for selecting
and contacting individual respondents from the
population.
|
|
|
|
sampling error | The degree to which the results
from the sample deviate from those that would be
obtained from the entire population, because of random
error in the selection of respondent and the
corresponding reduction in reliability.
|
|
|
|
scale interpretation error | Error associated with
the use of rating cards where respondents answer
with the name of a category, rather than its number
or code, and the interviewer records the wrong
code value because the category names are not
listed on the questionnaire.
|
|
|
|
scatterplot | A graphic plot of the data points for
two variables, usually generated on request by
analysis routines during regression analysis, so that
each data point is plotted horizontally according to
the value of the independent variable and vertically
according to the value of the dependent variable.
|
|
|
|
secondary data | Data initially acquired for general
use or for some purpose other than the information
requirements of the project at hand.
|
|
|
|
selection bias | A systematic effect on the data
resulting from the selection of respondents in a
manner that deviates from random selection, so
that some types of respondents are over- or underrepresented
in the sample.
|
|
|
|
self-selection bias | A systematic affect on survey
results because some respondents voluntarily
participate while others decline or refuse, so
that those with certain opinions or conditions
are under- or overrepresented in the sample.
|
|
|
|
semantic differential scale | A scaling device
that lists several pairs of bipolar adjectives, usually
separated by a seven-point scale, and instructs
respondents to rate the topic or object on each,
ordinarily used to measure image and provide
a profile.
|
|
|
|
semantic distance scale | A scale listing several
adjectives or phrases, where respondents are instructed
to indicate how well each describes some
object or topic based on a linear, numeric scale,
often used to measure image and obtain a profile in
much the same manner as the semantic differential
scale is used.
|
|
|
|
sequential sample | A sampling design that
requires the collection of data in increments with
a relatively small sample at each stage, so that
analysis can be performed after each stage to determine
when the sample is large enough to provide
the required level of confidence or reliability.
|
|
|
|
shape | The form or outline of a data distribution
arrayed from lowest to highest value, portrayed in
a data plot or described by statistical values such
as the coefficients of skewness and kurtosis.
|
|
|
|
sight-edit | The visual examination of the completed
questionnaires immediately after data
collection to determine if they are sufficiently
complete and usable.
|
|
|
|
significance level | The probability that the magnitude
of the relationship might result in a sample of
that size purely from sampling error if, in fact, it
did not exist in the population.
|
|
|
|
simple random sample | A sampling design that
seeks to select respondents from the population or
sample frame in a completely random fashion, so
every respondent has an equal probability of being
selected, and no clustering or stratification methods
are used.
|
|
|
|
single response item | A structured, multiplechoice
survey question that requires the
respondent to choose only one response
category, such as one that represents the "best"
or "favorite."
|
|
|
|
skewness | A designation of the shape of a distribution,
indicating the degree of symmetry or the
degree and direction that the mode or peak "leans"
toward one side, with only a few values extending
well out toward the tail on the other.
|
|
|
|
slope | In regression analysis, the "rise over the
run" when the dependent variable is plotted on
the vertical axis of a scatterplot, or the amount of
increase or decrease in the units of the dependent
variable for each unit of the independent variable,
indicated by the regression coefficient.
|
|
|
|
social desirability | The tendency for respondents
to give answers to survey questions that are consistent
with what the society believes is right, proper,
correct, or acceptable, creating bias in the results
whenever the true answers are suppressed to meet
social norms.
|
|
|
|
spread | The range and degree of dispersion or
variance in the distribution of data for a survey
variable, often described by the standard deviation
of the distribution.
|
|
|
|
stair-step scale | One type of pictorial scale graphically
showing the scale points as a series of steps,
appropriate for use with young children or other
respondents who might have difficulty understanding
a numeric or verbal scale.
|
|
|
|
standard deviation (S.D.) | A computed measure
of spread or dispersion in a distribution of data,
based on the squared deviations of each point
from the mean, that can be used to indicate the
proportion of data within certain ranges of scale
values when the distribution conforms closely to
the normal curve.
|
|
|
|
standard error (S.E.) of the estimate | A computed
value in regression analysis based on sample
size and variance around the regression line,
determining the confidence interval around
a predicted value of the dependent variable
at a given probability.
|
|
|
|
standard error (S.E.) of the mean | A computed
value based on the size of the sample and the standard
deviation of the distribution, indicating the
range within which the mean of the population is
likely to be from the mean of the sample at a given
level probability.
|
|
|
|
statistic | Some value computed from sample data
that may also be used to make inferences about the
corresponding value or "parameter" for the whole
population.
|
|
|
|
statistical analysis | The process of computation
and manipulation of sample data in order to
suppress the detail and make relevant facts and
relationships more visible and meaningful, and
to generate statistics in order to make inferences
about the population as a whole.
|
|
|
|
statistical inference | The process of generalizing
information from a sample to the population as a
whole by estimating population parameters, based
on their corresponding statistical values from
the sample.
|
|
|
|
statistical significance | An explicit assumption
by the analyst that a relationship revealed in the
sample data also exists in the population as a
whole, based on the relatively small probability
that it would result only from sampling error if
it did not exist in the population.
|
|
|
|
stratified sampling | A sampling design that
divides the population into specific strata containing
certain types of respondents, then selects
subsamples of the required size for each strata.
|
|
|
|
stratum | The singular form of "strata," indicating
one level of a stratified sampling design.
|
|
|
|
structured question | Any question that lists or
prescribes the response alternatives from which
respondents must choose, such as multiple-choice
or true/false questions and items accompanied by
rating scales.
|
|
|
|
subsample | One part of an entire sample
that is singled out for special attention or analysis,
often defined in terms of a demographic
characteristic.
|
|
|
|
sum | The total of a series of values or the process
of adding them.
|
|
|
|
sum of squares (S.S.) | A value computed for
several forms of statistical analysis, such as computing
the standard deviation, analysis of variance,
regression analysis, and the like, where some mean
is subtracted from each data point, this deviation
is squared, and the squared values are added for
all the cases.
|
|
|
|
survey | A research technique where information
requirements are specified, a population is identified,
a sample selected and systematically questioned,
and the results analyzed, generalized to the
population, and reported to meet the information
needs.
|
|
|
|
systematic | A relationship or effect that is not random,
but rather one that is consistent or in a given
"direction."
|
|
|
|
systematic bias | A redundant term, since bias is
defined as systematic effect, but commonly used
to emphasize the nonrandom nature of a bias or to
distinguish bias from random error.
|
|
|
|
systematic sampling | Another term for nth name
sampling, where the number of units in the sample
frame is first divided by the desired sample size to
obtain the value of n, a value between one
and n is
randomly selected as a starting point or first case
to be selected, and then every nth name or unit is
selected, yielding a random sample.
|
|
|
|
system hardware | Any mechanical or electronic
device linked in a computer system, including the
central processing unit and "peripheral" devices
such as printers and external disk drives.
|
|
|
|
t-distribution | A symmetrical statistical probability
distribution, slightly flatter and wider than the
standard, normal distribution, often listed in statistical
reference tables and used to determine the
significance of paired t-tests.
|
|
|
|
telephone interview | Interview data collection
using the telephone to contact respondents, as opposed
to personal interviewing where respondents
are in the presence of the interviewer and have visual
contact.
|
|
|
|
termination bias | Bias resulting when respondents
of a certain type or with a certain orientation
terminate their participation in a continuing
study, such as a panel study, at greater rates than
others.
|
|
|
|
threat | A source of bias or resistance resulting
when respondents find survey items or topics
intimidating or threatening, such as questions
about financial matters, or instructions, questions,
or scales that are confusing, suggesting the respondent
is ignorant or incompetent.
|
|
|
|
t-test | A statistical method of assessing the significance
of differences between two mean values
for the same variable, as opposed to a paired t-test
of values for two different variables for the same
cases, yielding the same basic information as an
analysis of variance with only two categories for
the independent variable.
|
|
|
|
Type I error | In academic or scientific research
(as opposed to pragmatic research), the probability
of rejecting the "null hypothesis" that no
relationship exists, and, therefore, accepting the
"alternative hypothesis" that there is a relationship,
when, in fact, no relationship exists in the population
as a whole.
|
|
|
|
Type II error | In academic or scientific research
(as opposed to pragmatic research), the probability
of not rejecting the "null hypothesis" that no
relationship exists, and, therefore, rejecting the
"alternative hypothesis" that there is a relationship,
when, in fact, a relationship does exist in the population
as a whole.
|
|
|
|
unaided recall | A form of questioning respondents
about what they remember, where the
facts, objects, or events are not listed or presented
to them to aid their recollection, as with aided
recall.
|
|
|
|
unbiased | Free of bias or unaffected by any
extraneous factor that would systematically
affect the values or results.
|
|
|
|
unbiased estimate | A statistical term that indicates
the value of a particular statistic, such as the mean,
obtained from the sample, will be exactly equal to
the corresponding value of the population parameter,
on the average over an infinite number of such
samples.
|
|
|
|
unconditional branching | Instructions that control
the "path" of interrogation by directing all respondents
who reach a specific place in the questionnaire,
such as the end of a special section, to go to
another location in the questionnaire, rather than
merely continuing from that point.
|
|
|
|
unimodal | Having only one modal value for
a distribution of categorical data or only one peak
or mode for a continuous distribution.
|
|
|
|
univariate analysis | The statistical description
or the analysis of just one variable at a time.
|
|
|
|
unstructured question | An "open-ended" survey
question where the alternative answers are not
listed and respondents must provide the answers
in their own words.
|
|
|
|
URL | Abbreviation of Uniform Resource Locator, the
global address of documents and other resources
on the World Wide Web.
|
|
|
|
validation | A term commonly but incorrectly used
by survey researchers and data collection agencies
to indicate "verification" of responses.
|
|
|
|
validity | The degree to which the survey data
or results are free from both systematic bias and
random error.
|
|
|
|
variable | A measurement unit that can be taken on
several different values, usually used to refer to the
distribution of data for one survey item.
|
|
|
|
variance | A statistical term referring to the sum
of the squared deviations of each data point from
the mean (the sum of squares), divided by the
number of cases or degrees of freedom (the mean
squares), and also the value from which the standard
deviation is computed by extracting the
square root.
|
|
|
|
verbal frequency scale | A particular type of
verbal scale where the frequency of an event to be
indicated by the respondent is expressed verbally,
ordinarily with the words "Always, Often, Sometimes,
Rarely, and Never," rather than in numeric
quantities.
|
|
|
|
verbal scale | Any scale whose points are
expressed in words or whose numeric code values
are labeled throughout the scale with words.
|
|
|
|
verification | The process of checking with respondents
after they have been interviewed to be sure
the person was actually interviewed and that the
interview was done correctly and completely when
and where it was supposed to be, and commonly
but incorrectly called "validation."
|
|
|
|
video-streaming | A new form of digital communication
sometimes used for focus group recording
and reporting, in which the video and audio are
simultaneously transmitted to a host server on the
Internet. The sponsor’s personnel authorized to
view the group are given the time schedule, sign
on instructions, and password, so they can observe
it from their own computer or conference room
whenever they wish.
|
|
|
|
visibility bias | One form of selection bias,
where a particular type of respondent is
over- or underrepresented in the sample because
they are more visible than others with different
characteristics.
|
|
|
|
yea-sayer | An individual or respondent who
persistently tends to respond in the affirmative
more often than others, regardless of the questions.
|
|
|
|
yea-sayer bias | The tendency for a set of survey
results to be generally and artificially positive on
a series of items because all items are inclined in the
same direction, toward the positive or toward the
negative, and positive responses to the earlier items
were generalized to the remaining ones, thus reducing
the validity.
|