This is a strange way to do this. In my experience, multiple imputation creates several data sets followed by a aggregated data set of those ones imputed. I would not run analyses on each one. I would run analyses on the aggregated one.

Cory

Hello Beatrice!

Although a multiple imputed dataset might tempt us into thinking that we may choose each one of these, this is not the correct approach to deal with multiple imputed datasets (as Cory mentioned). I'm not sure what software you used but if you used SPSS or R then you should obtain automatically pooled values (Although for SPSS, pooled values are restricted to some commands).

When using the MIXED command, you would obtain pooled results (for example for fixed effects, or if you use EMMEANS for post-hoc tests). You should directly interpret these instead of analysing your imputed data separately. What's also important is that pooled values are not obtained by simply creating an average of the values. There is a great online book that explains the multiple imputation procedure really well. It also explains Rubin's rule (the formula according which you would pool the results) without any super complicated formulas. It helped me a lot. https://bookdown.org/mwheymans/bookmi/multiple-imputation-models-for-multilevel-data.html

This discussion in the SPSS documentation website might be helpful:

https://www.ibm.com/support/knowledgecenter/en/SSLVMB_23.0.0/spss/mva/multiple_imputation_intro.html