Chelsea Hutto

PSYC 8000
Exam 3
50 points
Due 4/10/2014

1. Why is it important to determine the dimensionality of a measure before using it in

later analyses (regressions, ANOVAs, etc.)? (2 pt)
It is important to determine the dimensionality of a measure before using it in later
analyses in order to make sure scores are being used correctly in later analyses. For
example if there are multiple dimensions it is important to know the dimensionality so
there will be scores for each dimension.
Factor

1 2 3 4 5 6 7 8 9
2. In your own 0 words, explain what a latent variable is. (2 pt)
A latent variable is the overarching construct, we as researchers wish to measure.
2
Although this is somewhat misleading due to the fact we do not directly measure a latent
variable. We4 are only able to create measures, which seek to measure a particular latent
variable or 6construct.

3. Based on this
8 scree plot, how many factors should be extracted? (1 pt)

a. 4
b. 3
c. 2
d. 1

4. What does a factor loading tell us? (1 pt)

a. The strength of the factor
b. The magnitude of the relationship between a variable and a factor
c. The directionality of the relationship between a variable and a factor
d. Both b & c
e. All of the above

5. Oblique rotation should be used when: (1 pt)

a. You believe that the underlying factors are orthogonal.
b. You believe that the underlying factors are independent.
c. Kaiser’s criterion is met.
d. You believe that the underlying factors will be correlated.
 Chelsea Hutto

6. List and explain 2 differences between PCA and EFA. Name one situation in which
a PCA would be appropriate, and one situation in which an EFA would be
appropriate. (5 pts)
The purpose of these two measures vary greatly for example, a PCA is used to reduce the
number of items in a measure into smaller groups, which is essentially data reduction,
whereas an EFA is used to find latent constructs which are underlying the variables
measured in a test. EFA and PCA also differ in terms of their take on shared variance. A
PCA assumes the total variance of an item results from explained variance and error
which means shared variance is not due to an underlying factor, whereas an EFA assumes
for shared variance between two items are due to an underlying latent variable.
A situation in which a PCA could be used would be when a psychopathy scale that has
100 items, but would be more beneficial if the number of items was closer to 20.
A situation in which an EFA could be used would be when examining a construct that has
not been examined, or has not been examined fully. An example of such a construct
could include sociopathy.

7. What does the following output from a factor analysis tell us? (1 pt)

KMO and Bartlett's Test

Kaiser-Meyer-Olkin Measure of Sampling
Adequacy. .702

Bartlett's Test of Approx. Chi-Square 402.652

Sphericity df 153
Sig. .000

a. The sample size is adequate, and the correlations in the correlation matrix
are significantly bigger than zero.
b. The sample size is sufficient, but there is multicollinearity in the data.
c. The sample size is inadequate, and there is multicollinearity in the data.
d. The sample size is adequate, but the correlations in the correlation matrix are not
big enough.

8. Erik just used SPSS to find Cronbach’s alpha for a new scale of counterproductive
work behavior that he created. Because alpha is over .80, Erik decides that the scale
must be unidimensional. Is Erik correct? Give at least 2 reasons supporting your
answer. (3 pt)
Due to the value of alpha depending on the number of items on a scale, it is possible to
get a large a value on a scale due to the number of items rather than the actual reliability
of the scale. Secondly, alpha should not be used as an indicator of unidimensionality.
Although alpha is assumed to be an indicator of unidimensionality, this is only true when
there is one factor underlying the data. Additionally, reverse scored items will also affect
alpha, specifically, if an item is reverse phrased then it will have a negative relationship
with other items, therefore the covariances between this item and others will be negative.
 Chelsea Hutto

Based on this information, Erik should not conclude his scale is unidimensional due to
the value of alpha.

9. A reliability analysis was run on items from a questionnaire and gave rise to a
Cronbach’s alpha of α = –.8. What is the most likely explanation? (1 pt)

a. The questionnaire is unreliable.

b. The questionnaire has ‘inverse reliability’.
c. There were reverse-phrased items on the questionnaire that were not
reverse-scored before the analysis was done.
d. There are too few items on the questionnaire.

10. To determine which variables load on which factor(s), which matrix should be
interpreted? (1 pt)

a. Pattern matrix
b. Structure matrix
c. Factor matrix
d. Identity matrix

11. Esther wants to use an EFA to evaluate the dimensionality of her new scale of
narcissism. Based on previous work with narcissism scales, Esther has good reason
to believe that her data will not be multivariate normal. Which factor extraction
method should she use? (1 pt)

a. Principal axis factoring

b. Principal components analysis
c. Maximum likelihood
d. Oblique

12. Based on the following output, which item on this questionnaire would most
improve the reliability of the questionnaire if it were removed? (1 pt)
 Chelsea Hutto

R E L I A B I L I T Y A N A L Y S I S - S C A L E (A L P H A)

Item-total Statistics

Scale Scale Corrected

Mean Variance Item- Alpha
if Item if Item Total if Item
Deleted Deleted Correlation Deleted

Q1 54.1282 166.2200 .5505 .9082

Q2 54.5641 165.0418 .4917 .9092
Q3 56.0256 159.4467 .5719 .9072
Q4 54.6667 158.5439 .6972 .9039
Q5 55.0769 159.9676 .6225 .9058
Q6 54.9744 158.3414 .6422 .9052
Q7 54.6667 160.1228 .6374 .9055
Q8 54.5385 159.9393 .6314 .9056
Q9 55.1538 152.5020 .6596 .9048
Q10 55.6154 160.0850 .5321 .9084
Q12 56.0256 170.8151 .2423 .9156
Q14 54.3846 164.1377 .5508 .9078
Q16 54.9487 163.4710 .4447 .9108
Q17 55.0769 155.3887 .6760 .9041
Q18 55.5897 157.1430 .6421 .9051
Q19 56.0513 165.3131 .5756 .9076
Q20 55.0769 163.0729 .4870 . 9095
Q21 55.1026 152.9892 .7586 .9015

Reliability Coefficients

N of Cases = 39.0 N of Items = 18

Alpha =.9118

a. Q16
b. Q12
c. Q21
d. Q4

13. Comparable to linear models, a component in principal component analysis (PCA)

can be described as an equation. Which of the following equations is the correct
representation of a component in PCA? (1 pt)

a. Component i=b1 Variable 1 i+ b2 Variable 2i +…+ bn Variable ¿

b. Component i=b0 +b 1 Variable1 i+ b2 Variable 2 i+ …+bn Variable¿
 Chelsea Hutto

c. Component i=b0 +b 1 Variable1 i+ b2 Variable 2 i+ …+bn Variable¿ + ε i

d. All of these are incorrect because components in PCA cannot be represented as an
equation.

14. Which of the following sentences regarding the similarities of factor analysis and
PCA is false? (1 pt)

a. Both factor analysis and PCA are linear models in which loadings are used as
weights.
b. Both PCA and exploratory factor analysis take advantage of ‘rotation’.
c. Unlike PCA, factor analysis contains an error term.
d. In both PCA and factor analysis we predict components from measured variables.

15. Maura has just developed a new measure of emotional labor. She gave it to a group
of 80 employees working during the lunch rush at Chik-Fil-A. She wants to do an
EFA to examine the dimensionality of her new measure. Name and describe the 2
reasons why Maura should not use an EFA to analyze her data, and identify 1 thing
that Maura could do which would allow her to run an EFA to evaluate the
dimensionality of her measure. (3 pts)
Maura should not use an EFA to analyze her data for two reasons. The initial reason is
that her sample size is too small to conduct an EFA, which could jeopardize the results of
her study. Secondly, Maura chose to gather data from only one particular organization,
which limits the variability of the sample. Maura could gather data from another
organization, which could solve the problem of the sample size and variability.

16. An examiner of trainee chefs wanted to look at the underlying constructs of their 12
core culinary skills assessments. They ran a factor analysis with varimax rotation on
the scores of the 12 assessments. Looking at the rotated factor matrix from their
output below, how many variables load onto factor 1? (1 pt)
Factor
1 2 3
Chopping chillies 0.7 0.1 0.2
Cake decorating 0.1 0.8 0.3
Slicing an onion 0.8 –0.1 0.2
Combining colours on a plate 0.3 0.9 0.1
Cleaning and disinfection 0.2 0.0 0.9
De-seeding tomatoes 0.7 –0.1 0.0
Development of own recipes –0.1 0.8 –0.1
Ability to combine flavours 0.1 0.5 0.3
Keeping work surfaces clean 0.2 0.3 0.8
Skinning peppers 0.9 0.1 0.1
Personal hygiene 0.2 0.0 0.9
Knife skills 0.5 0.0 0.0

a. 11
 Chelsea Hutto

b. 6
c. 5
d. 1

17. An examiner of trainee chefs wanted to look at the underlying constructs of their 12
core culinary skills assessments. They ran a factor analysis with varimax rotation on
the scores of the 12 assessments. Looking at the rotated factor matrix from their
output below, which factor is the most likely to be labelled ‘preparation of fruit and
vegetables’? (1 pt)
Factor
1 2 3
Chopping chillies 0.7 0.1 0.2
Cake decorating 0.1 0.8 0.3
Slicing an onion 0.8 –0.1 0.2
Combining colours on a plate 0.3 0.9 0.1
Cleaning and disinfection 0.2 0.0 0.9
De-seeding tomatoes 0.7 –0.1 0.0
Development of own recipes –0.1 0.8 –0.1
Ability to combine flavours 0.1 0.5 0.3
Keeping work surfaces clean 0.2 0.3 0.8
Skinning peppers 0.9 0.1 0.1
Personal hygiene 0.2 0.0 0.9
Knife skills 0.5 0.0 0.0

a. Factor 1
b. Factor 2
c. Factor 3
d. None of the factors

18. A variable that shares none of its variance with any other variable would have a
communality of: (1 pt)
a. 1
b. 0
c. .5
d. 10

19. Name and describe 3 criteria that can be used to determine how many factors to
extract. (3 pts)
One criterion that can be used to determine the number of factors is Kaiser criterion.
Kaiser criterion retains all factors with eigenvalues of greater than one. A second
criterion is a scree test or plot, which graphical display of the eigenvalues. Thirdly,
 Chelsea Hutto

Substantive considerations are an additional way to determine the number of factors to

extract. This method uses past literature to determine the number of factors to extract.

20. Which of the following is true of factor scores? (1 pt)

a. Factor scores quantify the correlation between factors.

b. Factor scores can be used to assess how each variable correlates to each factor.
c. Factor scores can be used to overcome sphericity.
d. Factor scores can be used to overcome collinearity problems.

21. What is reliability? List and describe 3 ways in which reliability can be estimated.
(4 pts)
Reliability is the extent to which measurement is free from error. Which means if two
people with the same amount of a construct should get the same score on the
measurement. Reliability is a necessary condition for validity. One way reliability can be
estimated is through test-retest reliability which assuming all things being equal, a person
should get the same score on a questionnaire if they complete it at two different points in
time. A second way of estimating reliability is split half reliability, which splits a measure
in half and calculates scores for each half and then correlates scores on each half with the
other. Lastly, Cronbach’s alpha can be used to estimate reliability, which also overcomes
issues associated with split-half reliability. Cronbach’s alpha is also the most common
measure of scale reliability.

22. What are the 3 requirements for the existence of a cause-and-effect relationship
between 2 variables? (3 pts)
Inus condition which explains a cause in an insufficient but non-redundant part of an
unnecessary but sufficient condition for bringing about an effect. Secondly, a
counterfactual is something that is contrary to fact. Counterfactual is also what would
have happened if the IV had simultaneously not been manipulated. Lastly, the effect is
the difference between what did happen when the IV was manipulated and what would
have happened if the IV was not manipulated.

23. Which of the following is a non-manipulable cause? (1 pt)

a. Type of anxiety therapy
b. Severity of anxiety disorder
c. Gender
d. Both b & c

24. What is the primary difference between a randomized experiment and a quasi-
experiment? (1 pt)

In a randomized, participants are assigned to conditions be a random process, whereas in

a quasi participants are not randomly assigned to conditions.
 Chelsea Hutto

25. What is causal explanation? What are 2 analyses that can help with causal
explanation? (2 pts)
Causal explanation is determining why a causal mechanism works the way it does.
Both moderator variables and mediator variables can help with causal explanation.

26. In your own words: What is validity? What is a threat to validity? (2 pt)
Validity is a judgment of the extent to whether an inference is true or not. Things that
could serve as threats to validity include discrepancies in data acquisition or the
measurement process. For example, biased participants in a sample, or samples that focus
only on a particular group of individuals, or lack of random assignment.

27. Marlon conducted a study evaluating the effect of several different leadership-
training programs on subordinate ratings of leadership. Subordinates used an
ordinal scale to rate their supervisors’ leadership skills. Marlon used an ANOVA to
compare leadership ratings across training programs. Which threat to validity does
Marlon’s experiment contain? (1 pt)

a. Threat to statistical conclusion validity

b. Threat to external validity
c. Threat to construct validity
d. Threat to internal validity

28. Amy conducted a study to evaluate the effect of a new schizophrenia treatment
program on schizophrenia patients over a 10-year period. Individuals with the
highest scores on a psychoticism scale were chosen to participate in the treatment
group, while those with lower scores were chosen to participate in the control group.
Over the 10-year period, 10% of participants in the treatment group dropped out of
the study, and 25% in the control group dropped out. Name and describe 3 threats
to internal validity in Amy’s study. Name one way in which she could have
prevented one of these threats. (4 pts)

One of the threats to Amy’s internal validity is the participant used in the control group.
Rather than using participants that did not have significant levels of schizophrenia, Amy
used participants that scored lower on the psychoticism scale.
A second threat is attrition; due to the fact that not all participants in the study (or the
same proportion in each group) completed the experiment, this can affect the internal
validity of the measure.
A third threat is additive and interactive threats to internal validity. For example, the
groups could contain unequal numbers of men and women, and the levels of
schizophrenia are usually different for men and women.
One of the ways Amy could have eliminated the threats to internal validity, is by using
participants that did not have any levels of psychoticism rather than the use of
participants that scored lower.