I have a dataset of transactions done by consumers over 5 years. I would like to run a fixed effects regression on this dataset. What would be an appropriate unit of observation in this context? The issue is that each individual has more transactions per year.
If I choose the individual transactions as the unit of observations i get a pooled cross-section, with which I cannot use fixed effects.
Would the correct method be to select one transaction per person per year?