Psychological Assessment

Chapter 3 A Statistic Refresher specifically measured by the test

or measurement.
• Test scores are frequently
expressed as numbers, and • Nominal scales are the simplest
statistical tools are used to form of measurement. These
describe, make inferences from, scales involve classification or
and draw conclusions about categorization based on one or
numbers. more distinguishing
characteristics, where all things
• In this statistics refresher, we measured must be placed into
cover scales of measurement, mutually exclusive and
tabular and graphic presentations exhaustive categories.
of data, measures of central
tendency, measures of variability, • Ordinal scales permit
aspects of the normal curve, and classification. However, in
standard scores. addition to classification, rank
ordering on some characteristic is
• Measurement is the act of also permissible with ordinal
assigning numbers or symbols to scales.
characteristics of things (people,
events, whatever) according to • In business and organizational
rules. The rules used in assigning settings, job applicants may be
numbers are guidelines for rank-ordered according to their
representing the magnitude (or desirability for a position.
some other characteristic) of the
object being measured. • In clinical settings, people on a
waiting list for psychotherapy may
• Scale is a set of numbers (or be rank-ordered according to
other symbols) whose properties their need for treatment.
model the empirical properties of
the objects to which the numbers • Ordinal scales have no absolute
are assigned. zero point. In the case of a test
of job performance ability, every
• Thus, a scale used to measure a testtaker, regardless of standing
continuous variable might be on the test, is presumed to have
referred to as a continuous scale, some ability.
whereas a scale used to measure
a discrete variable might be • No testtaker is presumed to have
referred to as a discrete scale. zero ability.

• Measurement always involves • Zero is without meaning in such a

error. In the language of test because the number of units
assessment, error refers to the that separate one testtaker’s
collective influence of all the score from another’s is simply not
factors on a test score or known
measurement beyond those
 Psychological Assessment

• Interval scales contain equal In such an instance, the time

intervals. between numbers. taken to successfully complete
Each unit on the scale is exactly the puzzle is the measure that is
equal to any other unit on the recorded.
scale. But like ordinal scales,
interval scales contain no • A distribution may be defined as
absolute zero point. a set of test scores arrayed for
recording or study. The 25 scores
• With interval scales, we have in this distribution are referred to
reached a level of measurement as raw scores.
at which it is possible to average
a set of measurements and • As its name implies, a raw score
obtain a meaningful result. is a straightforward, unmodified
accounting of performance that is
• The difference in intellectual usually numerical.
ability represented by IQs of 80
and 100, for example, is thought • A raw score may reflect a
to be like that existing between simple tally, as in a number of
IQs of 100 and 120. items responded to correctly on
an achievement test.
• However, if an individual were to
achieve an IQ of 0 (something • In a frequency distribution, all
that is not even possible, given scores are listed alongside the
the way most intelligence tests number of times each score
are structured), that would not be occurred. The scores might be
an indication of zero (the total listed in tabular or graphic form.
absence of) intelligence.
• Often, a frequency distribution is
• Because interval scales contain referred to as a simple
no absolute zero point, a frequency distribution to
presumption inherent in their use indicate that individual scores
is that no testtaker possesses have been used and the data
none of the ability or trait (or have not been grouped.
whatever) being measured.
• In a grouped frequency
• Ratio scale. has a true zero distribution, test-score intervals,
point. All mathematical operations also called class intervals,
can meaningfully be performed replace the actual test scores.
because there exist equal
intervals between the numbers on • The number of class intervals
the scale as well as a true or used and the size or width of
absolute zero point. each class interval (or, the range
• example is a timed test of of test scores contained in each
perceptual-motor ability that class interval) are for the test
requires the testtaker to user to decide.
assemble a jigsaw-like puzzle.
 Psychological Assessment

• Graph is a diagram or chart

composed of lines, points, bars, • The Greek uppercase letter
or other symbols that describe sigma, Σ, is the symbol used to
and illustrate data. signify “sum”; if X represents a
test score, then the expression Σ
• With a good graph, the place of a X means “add all the test scores.”
single score in relation to a
distribution of test scores can be • The arithmetic mean The
understood easily. arithmetic mean, denoted by the
symbol X (and pronounced “X
• A histogram is a graph with bar”), is equal to the sum of the
vertical lines drawn at the true observations (or test scores, in
limits of each test score (or class this case) divided by the number
interval), forming a series of of observations.
contiguous rectangles.
• Symbolically written, the formula
• It is customary for the test scores for the arithmetic mean is X =
(either the single scores or the Σ(X/n), where n equals the
midpoints of the class intervals) number of observations or test
to be placed along the graph’s scores.
horizontal axis (also referred to
as the abscissa or X-axis) and for • The arithmetic mean is typically
numbers indicative of the the most appropriate measure of
frequency of occurrence to be central tendency for interval or
placed along the graph’s vertical ratio data when the distributions
axis (also referred to as the are believed to be approximately
ordinate or Y-axis). normal.

• A measure of central tendency • Shapes that frequency

is a statistic that indicates the distributions can take
average or midmost score • The median, defined as the
between the extreme scores in a middle score in a distribution, is
distribution. another commonly used measure
of central tendency. We
• Perhaps the most used measure determine the median of a
of central tendency is the distribution of scores by ordering
arithmetic mean (or, more the scores in a list by magnitude,
simply, mean), which is referred in either ascending or descending
to in everyday language as the order.
“average.”
• If the total number of scores
• Note that, in the formulas to ordered is an odd number, then
follow, the standard statistical the median will be the score that
shorthand called “summation is exactly in the middle, with
notation” (summation meaning one-half of the remaining scores
“the sum of”) is used. lying above it and the other half of
 Psychological Assessment

the remaining scores lying below • Some measures of variability

it. include the range, the
interquartile range, the semi-
• When the total number of scores interquartile range, the average
ordered is an even number, then deviation, the standard deviation,
the median can be calculated by and the variance.
determining the arithmetic mean
of the two middle scores. • The range of a distribution is
equal to the difference between
• These scores are said to have a the highest and the lowest
bimodal distribution because scores.
there are two scores (51 and 66)
that occur with the highest • The interquartile and semi-
frequency (of two). interquartile ranges A
distribution of test scores (or any
• Except with nominal data, the other data, for that matter) can
mode tends not to be a very be divided into four parts such
commonly used measure of that 25% of the test scores occur
central tendency. Unlike the in each quarter.
arithmetic mean, which must be
calculated, the value of the modal • the dividing points between the
score is not calculated; one four quarters in the distribution
simply counts and determines are the quartiles. There are three
which score occurs most of them, respectively labeled Q1,
frequently. Q2, and Q3. Note that quartile
refers to a specific point whereas
• Because the mode is arrived at in quarter refers to an interval.
this manner, the modal score
may be totally atypical—for • The interquartile range is a
instance, one at an extreme end measure of variability equal to the
of the distribution—which difference between Q3 and Q1.
nonetheless occurs with the Like the median, it is an ordinal
greatest frequency. In fact, it is statistic.
theoretically possible for a
bimodal distribution to have two • A related measure of variability is
modes. the semi-interquartile range,
which is equal to the interquartile
• Variability is an indication of how range divided by 2. Knowledge
scores in a distribution are of the relative distances of Q1
scattered or dispersed. and Q3 from Q2 (the median)
• Statistics that describe the provides the seasoned test
amount of variation in a interpreter with immediate
distribution are referred to as information as to the shape of the
measures of variability. distribution of scores.
 Psychological Assessment

• In a perfectly symmetrical
distribution, Q1 and Q3 will be the • Skewness is an indication of how
same distance from the median. the measurements in a
If these distances are unequal, distribution are distributed.
then there is a lack of symmetry.
• A distribution has a positive
• This lack of symmetry is referred skew when relatively few of the
to as skewness. scores fall at the high end of the
distribution. Positively skewed
• The average deviation Another examination results may indicate
tool that could be used to that the test was too difficult.
describe the amount of variability
in a distribution is the average • A distribution has a negative
deviation, or AD for short. skew when relatively few of the
scores fall at the low end of the
• The lowercase italic x in the distribution. Negatively skewed
formula signifies a score’s examination results may indicate
deviation from the mean. The that the test was too easy.
value of x is obtained by
subtracting the mean from the • The term testing professionals
score (X − mean = x). use to refer to the steepness of a
distribution in its center is
• The bars on each side of x kurtosis. To the root kurtic is
indicate that it is the absolute added to one of the prefixes
value of the deviation score platy-, lepto-, or meso- to
(ignoring the positive or negative describe the
sign and treating all deviation peakedness/flatness of three
scores as positive). All the general types of curves (Figure
deviation scores are then 3–6).
summed and divided by the total
number of scores (n) to arrive at • Distributions are generally
the average deviation. described as platykurtic (relatively
flat), leptokurtic (relatively
• The standard deviation as a peaked), or—somewhere in the
measure of variability equal to the middle—mesokurtic.
square root of the average
squared deviations about the • Distributions that have high
mean. kurtosis are characterized by a
high peak and “fatter” tails
• More succinctly, it is equal to the compared to a normal
square root of the variance. The distribution.
variance is equal to the • the normal curve is a bell-
arithmetic mean of the squares shaped, smooth, mathematically
of the differences between the defined curve that is highest at its
scores in a distribution and their center.
mean.
 Psychological Assessment

• From the center it tapers on both standard deviations above the

sides approaching the X-axis mean.
asymptotically (meaning that it
approaches, but never touches, • Researchers during World War II
the axis). developed a standard score with
a mean of 5 and a standard
• In theory, the distribution of the deviation of approximately 2.
normal curve ranges from Divided into nine units, the scale
negative infinity to positive was christened a stanine, a term
infinity. The curve is perfectly that was a contraction of the word
symmetrical, with no skewness. standard and nine.

• Standard score is a raw score • Standard score obtained by a

that has been converted from linear transformation is one that
one scale to another scale, where retains a. direct numerical
the latter scale has some relationship to the original raw
arbitrarily set mean and standard score.
deviation.
• The magnitude of differences
• Raw scores may be converted to between such standard scores
standard scores because exactly parallels the differences
standard scores are more easily between corresponding raw
interpretable than raw scores. scores. Sometimes scores may
With a standard score, the undergo more than one
position of a testtaker’s transformation.
performance relative to other
testtakers is readily apparent. • A nonlinear transformation may
be required when the data under
• If the scale used in the consideration are not normally
computation of z scores is called distributed yet comparisons with
a zero plus or minus one scale, normal distributions need to be
then the scale used in the made.
computation of T scores can be
called a fifty plus or minus ten • In a nonlinear transformation, the
scale; that is, a scale with a resulting standard score does not
mean set at 50 and a standard necessarily have a direct
deviation set at 10. numerical relationship to the
original, raw score.
• Devised by W. A. McCall (1922,
1939) and named a T score in • A coefficient of correlation (or
honor of his professor E. L. correlation coefficient) is a
Thorndike, this standard score number that provides us with an
system is composed of a scale index of the strength of the
that ranges from 5 standard relationship between two things.
deviations below the mean to 5
 Psychological Assessment

• An understanding of the concept

of correlation and an ability to • The coefficient of determination is
compute a coefficient of an indication of how much
correlation is therefore central to variance is shared by the X- and
the study of tests and the Y-variables.
measurement.
• There are many alternative ways
• Correlation is an expression of to derive a coefficient of
the degree and direction of correlation. One commonly used
correspondence between two alternative statistic is variously
things. A coefficient of correlation called a rank-order correlation
(r) expresses a linear relationship coefficient, a rank-difference
between two (and only two) correlation coefficient, or simply
variables, usually continuous in Spearman’s rho.
nature.
• Developed by Charles
• It reflects the degree of Spearman, a British psychologist
concomitant variation between this coefficient of correlation is
variable X and variable Y. The frequently used when the sample
coefficient of correlation is the size is small (fewer than 30 pairs
numerical index that expresses of measurements) and especially
this relationship: It tells us the when both sets of measurements
extent to which X and Y are “co- are in ordinal (or rank-order)
related.” form.

• The most widely used of all is the • Special tables are used to
Pearson r, also known as the determine whether an obtained
Pearson correlation coefficient rho coefficient is or is not
and the Pearson product- significant.
moment coefficient of
correlation. • One type of graphic
representation of correlation is
• Devised by Karl Pearson, r can referred to by many names,
be the statistical tool of choice including a bivariate
when the relationship between distribution, a scatter diagram,
the variables is linear and when a scattergram, or—our
the two variables being correlated favorite—a scatterplot.
are continuous (or they can
theoretically take any value). • A scatterplot is a simple
graphing of the coordinate
• The value obtained for the points for values of the X-
coefficient of correlation can be variable (placed along the graph’s
further interpreted by deriving horizontal axis) and the Y-
from it what is called a variable (placed along the graph’s
coefficient of determination, or vertical axis). Scatterplots are
r2. useful because they provide a
 Psychological Assessment

quick indication of the direction • A key advantage of meta-analysis

and magnitude of the over simply reporting a range of
relationship, if any, between the findings is that, in meta-analysis,
two variables. more weight can be given to
studies that have larger numbers
• Scatterplots are useful in of subjects.
revealing the presence of
curvilinearity in a relationship. • This weighting process results in
As you may have guessed, more accurate estimates (Hunter
curvilinearity in this context refers & Schmidt, 1990). Some
to an “eyeball gauge” of how advantages to meta-analyses
curved a graph is. are:
• (1) meta-analyses can be
• An outlier is an extremely replicated;
atypical point located at a • (2) the conclusions of meta-
relatively long distance—an analyses tend to be more reliable
outlying distance—from the rest and precise than the conclusions
of the coordinate points in a from single studies;
scatterplot. • (3) there is more focus on effect
size rather than statistical
• Outliers stimulate interpreters of significance alone; and
test data to speculate about the • (4) meta-analysis promotes
reason for the atypical score. evidence-based practice, which
may be defined as professional
• Meta-analysis may be defined as practice that is based on clinical
a family of techniques used to and research findings
statistically combine information •
across studies to produce single
estimates of the data under
study.

• The estimates derived, referred to

as effect size, may take several
different forms. In most meta-
analytic studies, effect size is
typically expressed as a
correlation coefficient.

• Meta-analysis facilitates the

drawing of conclusions and the
making of statements like, “the
typical therapy client is better off
than 75% of untreated
individuals”.