REPORTING EXPLORATORY FACTOR ANALYSIS’S RESULTS IN APA

Example 01
Table 01: Sample table for Exploratory Factor Analysis for Reasons to Watch TV
Loadings Communality
Items in the Scale Factor Factor
1: 2:
Watch TV for relief of boredom 0.774 0.599
Watch TV to escape from personal problems 0.744 0.564
Watch TV to cope with stress 0.728 0.575
Watch TV to change my mood 0.71 0.554
Watch TV to fill time 0.697 0.487
Watch TV for stimulation and excitement 0.646 0.433
Watch TV for things to talk about 0.574 0.551
Watch TV to socialize or spend time with other people 0.523 0.35
also watching
Watch TV to find out what’s going on in the world 0.815 0.665
Watch TV to keep in touch with what’s going on 0.756 0.572
Watch TV to challenge me, make me think 0.661 0.515
Watch TV for facts and opinion that will interest others 0.652 0.532
Watch TV to help me make a decision or learn something 0.65 0.429
Eigenvalue 3.894
% of Total Variance 29.953
Total Variance 52.50%

Twelve questions relating to reasons for watching TV were factor analyzed using principal
component analysis with Varimax (orthogonal) rotation. The analysis yielded two factors
explaining a total of 52.503% of the variance for the entire set of variables. Factor 1 was labeled
emotion/affective reasons to watch TV due to the high loadings by the following items: for
relieve of boredom; to escape from personal problems; to cope with stress; to change my mood;
to fill time; for stimulation and excitement; for things to talk about; to socialize or spend time
with other people also watching. This first factor explained 29.953% of the variance. The second
factor derived was labeled cognitive reasons to watch TV. This factor was labeled as such due to
the high loadings by the following factors: to find out what’s going on in the world; to keep in
touch with what’s going on; to challenge me, make me think; for facts and opinions that will
interest others; to help me make a decision or learn something. The variance explained by this
factor was 22.55%.
The communalities of the variables included are rather low overall with one variable (to socialize
or spend time with other people also watching) having a small amount of variance (35%) in
common with the other variables in the analysis. This may indicate that the variables chosen for
this analysis are only weakly related with each other. However the KMO and Bartlett’s Test of
Sphericity both indicate that the set of variables are at least adequately related for factor analysis.
Substantively, this means that we have identified two clear patterns of response among NCP03
respondents – one pattern of watching TV for emotional/affective reasons (or not), and one
pattern of watching TV for cognitive reasons (or not). These two tendencies are independent of
one another (i.e. they are not correlated).
Example 02
To further investigate the number of constructs and structure of this measure, an
exploratory factor analysis was conducted. Velicer’s Minimum Average Partial test and parallel
analysis were employed to determine the appropriate number of factors to retain (O’Connor,
2000). Although the sample size was small and unlikely to have enough power for an adequate
factor analysis, the analyses were done for the sake of comparison with the two-cluster solution.

The exploratory factor analysis using a principal-axis factor extraction was conducted to
determine the factor structure. Velicer’s MAP test recommended a three-factor solution, the
Parallel Analysis recommended a two-factor solution for the cyberloafing items, and the scree
plot indicated a two-factor solution. When comparing the two-factor solution to the three-factor
solution, double-loading of items was an issue in each solution, however, the two-factor solution
provided constructs that were less complex than those in the three-factor solution. For
interpretation of the two factors, a Varimax orthogonal rotation was used. This rotation had
sums of squared loadings ranging from 2.79 to 3.45, and the grouping of items was similar to
that of the cluster analysis where the first factor was behaviors frequently performed and the
second factor was behaviors infrequently performed (see Table 2). The frequent cyberloafing
behavior construct had a Cronbach’s alpha of .848, and the infrequent cyberloafing behavior
construct had a Cronbach’s alpha of .729, which could be increased to .752 if item 3 (“Browse
investment-related Web sites”) was removed. Although the factor analysis was calculated with
an insufficient sample size, and there was a large number of double-loaded items between the
two measures, it shows more evidence of a two-factor structure consisting of frequent and
infrequent cyberloafing behaviors.

Table 02: Factor Analysis Cyberloafing Constructs.

Prompt: How often do you engage in each activity during work hours for personal
reasons?
Scale: Never (0), Rarely (about once a month) (1), Sometimes (at least once a
week) (2), Frequently (at least once a day) (3)
 Factor
Loading
Item 1 2
Factor 1: Frequent Cyberloafing Behaviors ( = .848)
5. Browse general news Web sites .76 .09
6. Browse non-work-related Web sites .71 .04
8. Send non-work-related e-mail .68 .06
7. Check non-work-related e-mail .62 -.01
4. Browse entertainment-related Web sites .61 .30
9. Receive non-work-related e-mail .60 -.02
2. Shop online for personal goods .59 .23
18. Send or receive personal text messages .59 -.30
1. Browse sports-related Web sites .49 .25
15. Post messages on non-work-related items .47 .31
14. Chat with other people with instant messenger .41 .18
19. Make personal phone calls .39 .22
11. Download non-work-related information .28 .25
21. Use Facebook .27 .20
22. Use Instagram .23 .13
Factor 2: Infrequent Cyberloafing Behaviors ( = 729)
12. Download online games -.02 .79
10. Play online games .03 .67
17. Read or write in a blog .06 .63
23. Download Media (music / videos) .24 .60
13. Chat with other people in online chat rooms .21 .48
16. Use the Internet to gain additional income while at work -.02 .45
20. Use Twitter .32 .44
3. Browse investment-related Web sites .26 .31
Note. Double-loaded items are denoted in bold font.

SAMPLE MODEL FIT REPORT FOR CFA

Table 03: Model Validity Measures

CR AVE MSV MaxR(H) 1 2 3 4 5 6 7

1 0.927 0.716 0.603 0.932 0.718
2 0.950 0.705 0.557 0.961 0.288*** 0.840
3 0.931 0.691 0.298 0.934 0.427*** 0.484*** 0.831
4 0.942 0.731 0.557 0.945 0.326*** 0.746*** 0.545*** 0.855
5 0.869 0.528 0.050 0.879 0.086* 0.000 0.164*** 0.041 0.727
6 0.923 0.754 0.083 1.003 -0.259*** -0.104** -0.167*** -0.097** 0.013 0.868
7 0.931 0.603 0.303 0.937 -0.777*** -0.240*** -0.447*** -0.249*** -0.224*** 0.287*** 0.776

The model validity analysis

The model shows reliability as well as both discriminant and convergent validity. The values of are Composite Reliability (CR) above
0.9 indicating good reliability of the scales (Malhotra & Dash, 2011). The Average Variance Extracted (AVE) are greater than 0.7
indicating that there is convergent validity. Also, the Maximum Shared Variances (MSVs) are less than AVEs indicating discriminant
validity (Hair, Black, Babin, & Anderson, 2010).
Sample Report of Model-fit
The CFQ’s factor structure was tested through Confirmatory Factor Analysis (CFA), conducted
through IBM SPSS AMOS (version 20.0), aiming at confirming the one-dimensional structure of
the Cognitive Fusion Questionnaire found by its original authors. Maximum Likelihood (ML), a
widely used estimator in these statistical procedures (Brown 2006), was chosen as estimation
method. Global adjustment of the model was examined with the ChiSquare test (χ 2 ), the
Normed Chi-Square (NC), the Iterative Fit Index (IFI), the Tucker-Lewis Index (TLI), the
Comparative Fit Index (CFI), and the Root-Mean Square Error of Approximation (RMSEA), in
line with the original authors’ election. The following reference values were considered to assess
the goodness of fit of the models under investigation: a statistically non-significant Chi-Square
(Bollen 1989); the lowest NC possible or, at least, ranging between 2 to 5 (Bollen 1989;
Wheaton et al. 1977; Tabachnick and Fidell 2007); IFI, TLI and CFI above .95 (Brown 2006; Hu
and Bentler 1998); and RMSEA under .05 as indicative of an excellent model fit, ranging
between .05 and .08 as indicative of reasonable error and acceptable fit, between .08 and .10 as
indicative of mediocre fit, and above .10 as an unacceptable fit (Browne and Cudeck 1993; Hu
and Bentler 1998; Marôco 2014). Local adjustment indices, namely standardized regression
weights, squared multiple correlations and corrected itemtotal correlations, were checked
according to authors’ guidelines (e.g., Marôco 2014; Tabachnick and Fidell 2007). Internal
consistency of the instrument was assessed through Cronbach’s alpha coefficient, with values
above .70 considered to be acceptable according to Nunnally (1978). And paired-samples t-tests
were conducted to explore test-retest reliability in a subgroup of sample I. These specific
analyses were conducted with the IBM SPSS Statistics (version 20.0) software.
 REFERENCES
Hair, J., Black, W., Babin, B., and Anderson, R. (2010). Multivariate data analysis (7th ed.):
Prentice-Hall, Inc. Upper Saddle River, NJ, USA.
Hu, L., Bentler, P.M. (1999), "Cutoff Criteria for Fit Indexes in Covariance Structure Analysis:
Conventional Criteria Versus New Alternatives" SEM vol. 6(1), pp. 1-55.
Malhotra N. K., Dash S. (2011). Marketing Research an Applied Orientation. London: Pearson
Publishing.
Malhotra N. K., Dash S. argue that AVE is often too strict, and reliability can be established
through CR alone.
Malhotra N. K., Dash S. (2011). Marketing Research an Applied Orientation. London: Pearson
Publishing.
Pinto-Gouveia, J., Dinis, A., Gregório, S., & Pinto, A. M. (2018). Concurrent effects of different
psychological processes in the prediction of depressive symptoms–the role of cognitive
fusion. Current Psychology, 1-12.