8
$\begingroup$

Context

I have got some problems with my doctoral dissertation. My thesis is Investigating Secondary Primary School Teachers' Organizational citizenship behaviours through their perceptions about organizational culture and their organizational trust levels.

I have a sample of 871 teachers. I have three instruments but they were developed by other researchers and they were used in some other studies.

I have been trying to analyse my data using structural equation modelling. However while trying to do confirmatory factor analysis, only one instrument (organizational citizenship) was ok. The other two intruments did not give the first order confirmatory analysis. The RMSEA values were around 0.100. Chi-square was too high, and chi-square divided by degrees of freedom was also too high.

Question

  • What should I do with my instruments and CFAs?
  • Would item parcelling or removing items from these instruments be a good idea?
  • Or should I skip structural equation modelling and go on with regression?
$\endgroup$
5
$\begingroup$

I would probably do the following: 1) Split the data into two roughly equal segments. 2) Perform exploratory analyses on one of these and derive a new model 3) Test the model on the other half of the data.

This at least will be something that's not done all that often, which will make it a better fit for publication (should you want to do so), and will give you an independent test of your model.

You could also fit both models (the prior one and the one you develop) to your test data, and compare the fit of both.

$\endgroup$
1
$\begingroup$

Instead of looking for statistical solutions that directly solve this problem, I would look for solutions that improve the diagnosis.

First, I'd compare the different samples used in the different studies.

Then, if you have the data, I'd look at the correlation patterns among the variables in the different samples. (You may be able to get these from other authors).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy