For my pilot study, i have collected 28 responses. As i have small data set, so i will have to use SPSS to analyse the data. However, i have second order constructs. I have performed factor analysis once to get three constructs from various items. I have to make one construct from these 3 constructs. This way, i have to form 3 constructs from 9 constructs.
How should i do that? Your suggestions is appreciated.
Regards,
Abdul
"As i have small data set, so i will have to use SPSS to analyse the data. " What's the relation between using a software and sample size?
I get that it's a pilot study, but 28 responses is very small for you to have results you can trust before going further. Trying going up to 50-100, especially for factor analysis, and especially if the structure is complex (more latent constructs to estimate needs more data to estimate them, basically).
Anyway, here's something more practical, see answers to this question on RG.
https://www.researchgate.net/post/How_can_I_perform_second_order_factor_analysis_in_SPSS/1
Rule of thumb is that you need at least 100 subjects to conduct EFA or five subjects per variable - especially if you have latent constructs (see Gorsuch). Hope this helps
Also Are you trying to build a Path Model ?
SmartPLS will help you in this case, in case of less data points. Also Its easy to build 2nd order constructs...
With that sample size, try with Partial least squares -SEM (PLS-SEM).
Abdul,
While I concur with others that a bigger sample size would be preferred, let me answer your question.
You have two options using SPSS; each presumes that you used an oblique rotation (e.g., Promax or Oblimin) for final factor solution:
1. WIthin the SPSS factor analysis, there is the option to create and save factor scores for each case to your worksheet. Do this. Then, run a second factor analysis, using the three factor scores as the variables to be factored.
2. Enter the summary correlations among the three factors using syntax similar to that below (for a 3-variable problem), then run the factor subprogram (again, using syntax).
If you have the AMOS add-on, you can directly specify higher-order factors, but the default context is CFA, not EFA.
Good luck with your study!
David Morse
MATRIX DATA VARIABLES = x1 x2 x3
/FILE= INLINE
/FORMAT= LOWER DIAG
/CONTENTS=corr N .
begin data
1.0
0.64 1.0
0.48 0.48 1.0
100 100 100
end data.
factor
/matrix = in(cor = *)
/ extraction = ml
/ rotation = oblimin
/ print = default correlation kmo aic repr
/ plot = eigen .
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS