PSY 7 PSYCHOLOGICAL ASSESSMENT 1 (TEST CONSTRUCTION AND DEVELOPMENT)
1 Steps For Doing Exploratory Factor Analysis And Confirmatory Factor Analysis Using
Statistica
EXPLORATORY FACTORY ANALYSIS
STEPS:
I. DATA CLEANING (make sure your excel file does not contain any character
like comma or single quotation mark except the numbers)
1. Statistica
2. Open file
3. Import file
II. DATA TRANSFORMATION
If the distribution of your data is skewed, use the appropriate transformation
method:
SKEWNESS VALUES TRANSFORMATION METHOD
Moderately positive skewness Square-Root
(need not transform) NEWX = SQRT(X)
Substantially positive skewness Logarithmic (Log10)
NEWX = LG10(X)
Substantially positive skewness (with zero Logarithmic (Log10)
values) NEWX = LG10 (X+C)
Moderately negative skewness SQUARE-Root
NEWX = SQRT (K-X)
Substantially negative skewness Logarithmic (Log10)
NEWX = LG10 (K-X)
X = refers to the value of the data for a participant.
For example, if Participant’s answer is 3 on item 1 (and item 1 is skewed),
use 3 as your X. Use the appropriate transformation method formula
C = a constant added to each score so that the smallest score is 1.
K = a constant from which a score is subtracted so that the smallest score is 1:
usually equal to the largest score +1.
For example, from a scale of 0 to 4, largest score is 4. Using the formula
above, K = 4 + 1, or K = 5
1 To check for skewness, click on Descriptive Statistics, Skewness.
Transform individual data per item with
Source:
File: ydelapena | department of psychology - XU
PSY 7 PSYCHOLOGICAL ASSESSMENT 1 (TEST CONSTRUCTION AND DEVELOPMENT)
2 Steps For Doing Exploratory Factor Analysis And Confirmatory Factor Analysis Using
Statistica
[Link]
n%[Link]
Note: Statistica rule of thumb
skewness is between +2.0 and -2.0.
+1 or -1 > is already moderately skewed.
0.5+/- is symmetrical. The more one transforms the data, the more it no longer
approximates actual data.
III. EXPLORATORY FACTOR ANALYSIS
What to click:
1. Statistics
2. Multivariate
3. Factor analysis
4. Click all variables
5. Ok
6. Set maximum no. of factors based on assumption; eigenvalue set at 1
7. Ok
8. Maximum Likelihood when you are sure that you have a Multivariate
Normality (MAHALANOBIS in SPSS to check for multivariate normality.
Skewness is only of Univariate normality). Use Principal Axis Method is not
particular or strict on normality issue.
9. Varimax raw (rotates all item to one another)
10. Factor loading set at .40 (.70 is so stringent)
11. Loadings
12. Can check first Eigenvalues. Look at Cumulative %. But the more factor
loadings added, the Cumulative % also increases. Could consider scree
plot.
13. Summary
14. Extraction: Factor analysis, explained variance, eigenvalues
15. Factor analysis result, Scree Plot.
Consider the start of the elbow – point of inflection. Where X and Y
axis meets
Theoretically, minimum number of items per factor is at most 3.
2
Could consider 2 items.
Less than 1 eigenvalue is no longer processed by Statistica. By
default Statistca sets it at 1.
File: ydelapena | department of psychology - XU
PSY 7 PSYCHOLOGICAL ASSESSMENT 1 (TEST CONSTRUCTION AND DEVELOPMENT)
3 Steps For Doing Exploratory Factor Analysis And Confirmatory Factor Analysis Using
Statistica
If one should run CFA and factor loading is questionable, go back to
Reliability Analysis. It will check consistency of items in a scale. It
also identifies items to delete. Afterwards, redo CFA afterwards
Option 1: Conduct EFA first before checking for Cronbach Alpha.
Option 2: Could start first with Item Analysis before doing EFA to
check for intercorrelations. This will help you determine which items
to drop prior to EFA. After EFA, run again the Item Analysis and
check for potential items you can drop.
Coefficient Alpha boundary is 0.7. Check only inter-item correlation
with coefficient alpha is really low. Otherwise, could retain items
when Cronbach alpha is sufficient.
CONFIRMATORY FACTOR ANALYSIS
When to use: to check if existing theory that specifies the dimensions are
confirmed by your data. There is an existing theory that specifies how many
dimensions are logically and systematically involved in a model.
Create another questionnaire (now containing the items identified per factor)
and then do CFA.
STEPS:
When you’re done with your EFA and already performed the 2nd test administration and
encoded in Excel file, make sure you change the heading or title for each item such that it
reflects the factor you have considered. For example, if you named items 1 to 10 as
Humor factor, label items as H1, H2, H3 to H10. If factor 2 is named Kindness with 5
items, label items as K1, K2, to K5.
I. DATA CLEANING (make sure your excel file does not contain any character
like comma or single quotation mark except the numbers)
4. Statistica
5. Open file
6. Import file
3 II. CFA steps
1. Statistics
2. Advanced linear
3. Structural equation model
File: ydelapena | department of psychology - XU
PSY 7 PSYCHOLOGICAL ASSESSMENT 1 (TEST CONSTRUCTION AND DEVELOPMENT)
4 Steps For Doing Exploratory Factor Analysis And Confirmatory Factor Analysis Using
Statistica
4. Path wizard
5. Confirmatory factor analysis
6. Ok
7. Select variables per factor. Click VAR. Select items per factor. OK. Type in
name per factor.
8. Ok
9. Run model
10. Check if converging normally. If not converging normally, transform data.
11. Ok
12. Click Basic Summary Statistics
13. Divide ML chi-square over degrees of freedom. Check if less than 3. Less
than 3 is valid. (meaning the instrument or structure)
14. Structural equation model for Noncentrality (RMSEA). Check for validity
index. At least 1 validity from the 4 validity indeces. (# 13 as the fourth
validity index)
Look at RMSEA : Steiger’s value. Standard is less than .08.
Click Other single: GFI must be greater than 0.9.
NFI should be more than 0.9
RMSEA
File: ydelapena | department of psychology - XU