Hello,
I have a question about Mixed Models in SPSS.
I have a dataset with many different information about more than two thousand firms in Turkey for two years; 2008 and 2013. I'm testing the effect of the business climate obstacles on firm operations, and I'm using Mixed Models in SPSS due to the repeated measures from the two years.
Each firm has only one person answering the different questions (so only one entry for each firm). The problem is that not all firms have entries in both 2008 and 2013. Most of them have in either year only. For example, I have 998 unique entries for 2008, 949 unique entries for 2013, and only 97 firms that are repeated in both 2008 and 2013. Does it make sense to use mixed models or even to do a panel data analysis for repeated measures when my actual repeated measures are very low compared to the unique entries of each year? If not, what would be the alternative?
I would appreciate any input on this. Thank you.
You will know more about you data than I will but such a big difference in the cases between the two years suggests that they are not really compatible samples. Mixed models do allow missing outcomes under the forgiving MAR assumption and not the more demanding MCAR. But you will know if the missingness is likely to be informative (NMAR) and therefore a problem.
More helpfully you can take those that are there in both years and do a standard regression - there is no need for a mixed model with only 2 years. When publishing you will have to discuss / argue for what these cases are representative of.
I would put score Now as the response and score Past as a predictor variable so you are modelling change and the include other variables in the standard regression model.
Have a look at my responses to this
https://www.researchgate.net/post/What_are_the_statistics_for_a_cross_sectional_study
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS