Test And Measures Test 1

Front 1

Assumptions of Traditional Assessment

Back 1

1. Behavior is a function of underlying mental constructs
2. Behavior and constructs inferred from behavior are fundamentally different
3. 5 problems of psychological measures
4. Behavior signs are independent of the situation

Front 2

5 problems of psychological measurement

Back 2

1. No single approach universally accepted
2. Measurements based on limited sample
3. Measurement is subject to error
4. Measurement scales lack well defined units
5. Constructs are defined by 2 statistical relations relations

Front 3

Goals of Traditional Assessment

Back 3

1. Assign construct based on variability in performance
2. Explain / predict behavior from construct value
3. Select treatments that are effective for individuals with similar construct values

Front 4

Scaling Person's Score

Back 4

1. Compute mean as reference point
2. Compute deviation score x = X - u
3. Compute variance as index of total variability in set of scores
4. Compute z score as a standard scale
z = (x-u)/ o
5. Correlation: pxy = Sum or zxzy / N
6. Test scores as composite Scores

Front 5

Item discrimination Indices

Back 5

1. Total score as criterion
2. Item difficulty
3. Distribution of Responses
4. D Coefficient
5. Item total correlation
6. Item reliability/validity index

Front 6

Generating initial item pool

Back 6

1. Purpose
2. Identify behaviors represent construct
3. Table of Specifications
4. Initial item pool

Front 7

Vagonotic Measurement

Back 7

phenomena defined into existence based on variation in a set of underlying observations

Front 8

Mental constructs

Back 8

psychological property/ attribute; innate, stable, influence behavior in dimension

Front 9

Variance is the basis of

Back 9

Correlation

Front 10

Reliability

Back 10

degree to which score difference can be attributed to systematic sources of variation

Front 11

Systematic Variation

Back 11

know influences; non-random fluctuations

Front 12

True Score Theory

Back 12

the average of the observed scores obtained over an infinite # of repeated testings with same test.

Front 13

Realization of random variable

Back 13

when you have a distribution of scores you can see

Front 14

Expected value of random variable

Back 14

mean of observed scores = true score

Front 15

Reliability coefficient

Back 15

Correlation between true and observed scores; estimate correlation on parallel tests

Front 16

Standard error of measurement

Back 16

discrepancy between examinees observed test score and true score; SD one person's scores over k # of parallel tests; substitue SD for the total test to estimate SEM for 1 person; confidence intervals

Front 17

Generalizability Theory

Back 17

1. Estimates of reliability based on how test is going to be used
2. Identify sources of measurement error (of concern) and conduct reliability study to assess effects simultaneously

Partition garbage can error variance into different sources

Based on ratio of variance estimates from diff sources; get variance by deviating scores from grand mean

Front 18

Facet
Single Facet design

Back 18

grouping variable in ANOVA
one way anova

Front 19

Factor Analysis

Back 19

data reduction technique that use correlations among large amount of variables to create composite variables/factors

Looking for % of VAC ranks

Front 20

Matrix

Back 20

table of scores with columns of variables and rows of persons

Front 21

Factor

Back 21

A linear combination of variables in matrix

Front 22

Factor loadings

Back 22

correlation with an item with a factor

Front 23

Vector

Vector space

Hyperspace

Back 23

Each factor column can be considered a vector

2 or more

more than 2 dimensions of axis

Front 24

Decision 1

Back 24

Define factors

Principal components; total variance
common factors; covariance only
communality estimates; ult-r sq in diag on correlation matrix

Front 25

Decision 2

Back 25

Extracting Successive Factors:

Eigenvalue = sum of squared factor loadings for each factor or column

EIGENVALUE/ # of items = % VAC by a factor

MINEGAN = 1 = atleast as much % VAC as an item

Communality = proportion of an item's variance accounted for by the factors

Front 26

Decision 3

Back 26

Rotation of initial factors to teminal factors

VARIMAX - orthogonal = 90 degrees
QUARTIMIN - oblique = < 90 degrees

Front 27

VARIMAX

QUARTIMIN

Back 27

90

< 90

Front 28

Decision 4

Back 28

Interpreting Factors
Minimum factor loading for inclusion
Factor label - dimension represented

Front 29

Deviation Score

Back 29

x = X-u

Front 30

Variance

Back 30

=

Front 31

z score

Back 31

z = x-u/o

Front 32

correlation

Back 32

xy =

Front 33

variance of composite

Back 33

=

Front 34

Cronbach's Alpha

Back 34

= k

Front 35

SEM

Back 35

SEM = SD

Front 36

Generalizability coefficient

Back 36

persons

Front 37

EIGENVALUE

Back 37

sum of squared factor loadings for each column or factor