Basic Concepts in
Measurement & Statistics
0
Defining Measurement
• Psychological measurement is the process
of assigning numbers (that is, test scores)
to people.
• Scales of measurement;
• Nominal scales
• Ordinal scales
• Interval scales
• Ratio scales
0
Scales of Measurement
• Nominal scales (in name only)
• (e.g., eye color, gender, ethnicity)
• Ordinal scales (ordering)
• (e.g., ranking according to height)
• Interval scales (equal distances)
• (e.g., calendar years)
• Ratio scales (absolute zero)
• (e.g., weight in pounds, age
0
Scales of Measurement
0
Scales of Measurement
0
Distribution
• A distribution is a set of scores
Normal distribution: A theoretical distribution
with a symmetrical shape and the highest
frequency concentrated in the middle.
0
Distribution
0
Distribution
• A distribution is a set of scores
0
Distribution
Negatively Skewed. Positively Skewed.
The tail extends towards lower The tail extends towards higher
(negative) values. (positive) values.
0
Distribution
0
Distribution
Positively Skewed.
0
Evaluating Psychological Tests
• Reliability: A reliable test yields consistent scores when a
examinee takes two alternate forms of the test or when
s/he takes the same test on two or more different
occasions.
0
Evaluating Psychological Tests
• Validity
• Validity of measurement
• Whether the test adequately measures what it purports to measure.
• If this is the case, an intelligent person should receive higher
scores than do less intelligent people.
• Validity of decisions
• A valid test is useful for making accurate decisions about
individuals.
0
Statistical concepts
• 3 major concepts:
• Variability
• Allows us to measure and describe the extent to which test scores
differ.
• Computing the difference between each person’s score and the
mean.
• A large variance indicates that individual scores often differ from
the mean.
0
Variability
0
Statistical concepts
• Variability
• One application of the standard deviation is to form standard
scores, z scores.
• A (+) z score means that you are above the mean.
score
mean=2.6
frequency Z scores
4 4
sd=.94
3 6
2 8
1 2
0
Statistical concepts
• Correlation
• are often illustrated by using scatterplot.
Aptitude scores and grades
Correlation coefficient
describes the strength and direction of a
relationship between variables.
• a[
0
Correlation
0
Correlation
0
Statistical concepts
• r = .90 Strong, positive correlation
• r = - .85 Strong, negative correlation
• r = .21 Weak, positive correlation
• r = -.10 Weak, negative correlation
• Pearson Product-Moment Correlation Coefficient
• When variables are on an interval or ratio scale.
• Spearman Rank Correlation Coefficient
• When the variables are on an ordinal scale
• Point-Biserial Correlation Coefficient
• One variable dichotomous; one on an interval or ratio
scale 0
Statistical concepts
• Prediction
• We can predict one’s behavior from test scores.
0
Prediction
• Linear Regression allows you to predict values
on one variable given information on another
variable.
Note: Y = a + bX when a = 10 and b
= 0.5. For example, if X is 30, then Y
= 10 + (0.5)30 = 25.
0
Prediction
Percentile
• The point below which a specified percentage
of the observations fall.
• If a student’s IQ score (130, z= 2) is at 98th
percentile, 98% of the IQ scores are below
this score. Student’s score is better than 98%
of the all students.
0
Percentile
0
Percentile
0
Scales & Transformations
Comparing scales
A score of 4 on a 5-point scale.
What is the equivalent score of it on a 7-point scale?
0
Scales & Transformations
• Raw scores do not tell whether the subject did
well.
• Need more interpretable scores
• Characteristics of transformations:
• Doesn’t change a person’s score, just expresses it in
a different way.
• Takes into account information not contained in the
raw scores itself.
• Expresses the scores in more interpretable units.
0
Linear Transformations
• Changing number(s) by adding (+),
substracting (-), multiplying (x) or dividing (/)
• Transformed score= constant + (weight x raw
score)
• Most familiar linear transformation is z score
• Z=(X-M)/SD
• Easy to interpret (- and + scores), can be easily
converted to percentiles.
• Negative scores, have fractional values (at least 2
decimal points).
0
Linear Transformations
• z score
• A z score of 0?
• 84.1%?
0
Linear Transformations
0
Linear Transformations
• t score=(z score x 10)+50
score frequency z scores t scores
4 4 1.49
3 6 .43
2 8 -.64
1 2 -1.70
0
Area Transformations
• express a person's score in terms of where it
falls on the normal curve, rather than simply
providing a new unit of measurement (like
linear one did).
0
Area Transformations
• Percentile Scores
1. Cum Fm= (0.5 x f) + Cum F below
2. Percentile scores= Cum Fm x (100 / n)
score freq Cum f Cum Fm Percentile scores
4 4 20 .5 x 4 + 16 = 18 18x(100/20) = 90
3 6 16 .5 x 6 + 10 = 13 13x 5 = 65
2 8 10 .5 x 8 + 2 = 6 6x5 = 30
1 2 2 .5 x 2 + 0 =1 1x5 = 5
0
Area Transformations