Dear Stata users,

I have a query about the number of observations using the blow-up and cluster (buc) model.

Using fixed effect (fe) and blow up and cluster (buc) on a model to analyse parents’ influence on children’s snack (panel data), the number of observation across both models is consistently different, with buc having larger observations (e.g. fe n=6873 whilst buc: n=9000). Both models control for the same independent variables.

Note that the Stata syntax that I use does have “preserve” and ”restore” to ensure the data are returned to their original form after the blow up process.

I understand that the buc uses all variation of the ordinal response variable by expanding the data set to accommodate all possible binary recoding options of the ordered dependent variable. Could it be the reason why the number of observation is larger in buc in comparison with fe? Overall, should I expect more observation in a buc in comparison to fe?

Many thanks for your help.
​​​​​​​Zoe Ejebu