Configuring a longitudinal CFA model for pre/post-term survey analysis?

CFA
1

I’m working on a project where I need to compare factor models from pre and post-term surveys using longitudinal CFA. My dataset includes 12 questions grouped into 3 latent variables:

  • Motivation: Q1, Q2, Q3, Q4
  • Confidence: Q5, Q6, Q7, Q8
  • Performance: Q9, Q10, Q11, Q12

I’m not sure how to set up the configural model. Should I:

  1. Model each factor separately for pre and post-term?
  2. Combine all factors in one model?
  3. Include additional parameters?

Here’s a sample of what I’m thinking:

model <- '
Motivation.pre =~ Q1.pre + Q2.pre + Q3.pre + Q4.pre
Motivation.post =~ Q1.post + Q2.post + Q3.post + Q4.post
'

Also, can I use the equaltestMI package for this? Or is that only for multigroup CFA?

Any advice on the best approach would be really helpful. Thanks!

2

hey, i think combining all factors in one model is the way to go. it’ll give you a better picture of how everything’s connected over time.

something like this might work:

model <- '
Motivation.pre =~ Q1.pre + Q2.pre + Q3.pre + Q4.pre
Confidence.pre =~ Q5.pre + Q6.pre + Q7.pre + Q8.pre
Performance.pre =~ Q9.pre + Q10.pre + Q11.pre + Q12.pre
# same for post
'

dont forget to add correlations between pre and post factors too!

3

Great question! For your longitudinal CFA model, I’d recommend combining all factors in one model. This approach allows you to examine the relationships between factors over time and provides a more comprehensive view of your data structure.

Here’s a suggested model setup:

model <- '
# Pre-term factors
Motivation.pre =~ Q1.pre + Q2.pre + Q3.pre + Q4.pre
Confidence.pre =~ Q5.pre + Q6.pre + Q7.pre + Q8.pre
Performance.pre =~ Q9.pre + Q10.pre + Q11.pre + Q12.pre

# Post-term factors
Motivation.post =~ Q1.post + Q2.post + Q3.post + Q4.post
Confidence.post =~ Q5.post + Q6.post + Q7.post + Q8.post
Performance.post =~ Q9.post + Q10.post + Q11.post + Q12.post

# Correlations between pre and post factors
Motivation.pre ~~ Motivation.post
Confidence.pre ~~ Confidence.post
Performance.pre ~~ Performance.post
'

This model allows for factor correlations across time points, which is crucial for longitudinal analysis. Regarding equaltestMI, it’s primarily designed for multi-group CFA, but you can use lavaan for your longitudinal CFA. Consider using measurement invariance tests to ensure your factors are comparable across time points.

4

Hey there! I’m really intrigued by your longitudinal CFA project. It sounds like you’re diving deep into some fascinating pre/post-term survey analysis!

Have you considered adding autoregressive paths to your model? Something like this could be super interesting:

model <- '
# Your existing factor structure
Motivation.pre =~ Q1.pre + Q2.pre + Q3.pre + Q4.pre
Confidence.pre =~ Q5.pre + Q6.pre + Q7.pre + Q8.pre
Performance.pre =~ Q9.pre + Q10.pre + Q11.pre + Q12.pre
Motivation.post =~ Q1.post + Q2.post + Q3.post + Q4.post
Confidence.post =~ Q5.post + Q6.post + Q7.post + Q8.post
Performance.post =~ Q9.post + Q10.post + Q11.post + Q12.post

# Autoregressive paths
Motivation.post ~ Motivation.pre
Confidence.post ~ Confidence.pre
Performance.post ~ Performance.pre
'

This way, you could see how each factor at pre-term predicts itself at post-term. Pretty cool, right?

Oh, and have you thought about cross-lagged effects? Like, does pre-term motivation predict post-term performance? That could lead to some really juicy insights!

What software are you using for this, by the way? Lavaan? Mplus? I’m always curious about different tools for these kinds of analyses. :nerd_face: