I would like to ask for advice on data weighting - i.e. whether or not I should use post-stratification weights in my analyses? I have data collected by quota sampling, which are weighted with post-stratification weights to accurately represent the target population. Due to missing values, I am working with a smaller sample. I wanted to ask if I should do the missing value analysis and then further analysis on the smaller sample on the weighted data? Among the specific analyses, I am applying descriptive statistics, reliability testing, CFA and MGCFA.
I have tried both (weighted x unweighted) and there is not much difference in the case of descriptive statistics, but in CFA and MGCFA the results come out relatively quite different.
Thanks in advance for any advice and tips on how I should best proceed.
Obviously, given that factor analysis also includes the variables defined as latent and the related estimates.
Therefore, the absence or presence of weights can modify the results, even substantially.
It all depends on the purpose of your research.
Since the results from factor analysis depend entirely on the correlations among the variables, I suggest you look at the two sets of correlation matrices to get a sense of why the weighting leads to different results.
- Badges
- Science topic
- Similar topics
- Econ
- Innovation Economics
- Innovation