xtprobit regression with a cluster

Hello everybody,

I have a huge problem that I hope you can help me to solve, since I have not found anything helpful in the web.

I have a WIDE dataset, made of:
- a variabile called "people" (with 660names)
- 32 variables, called "action1".."action32", with value = 1 or = 0 (in order to have a probit)
- 1 dummy "gender"
- 32 variables "income" that change (the first 20 have a value, the second 8 another, the third 4 another, the last two another)
- 32 variables "motive", they are referred to the "action" variables (for every "action" I have a different "motive").

Now what I should do is to reshape the dataset into LONG, but if the i variable is people, the j variable should not go from 1 to 32, but from 1 to 4. Infact the actions are referred to 4 different "days" (the distinction operated for "motive": the first day has 20 actions, the second 8, and so on).
I was suggested to cluster the "actions", in order to have a dataset long with in order: "people", "days", "actions", the value of "action", "motive" and gender.
x1 1 1 0 n 1
x1 1 .. .. .. 1
x1 1 20 1 n20 1
x1 2 1 0 n21 1

and so on...

My problem is that, first, I know how to reshape a wide dataset into a long one, but when j is clear (e.g. I will take the 32 "actions" as a j); but now I need another passage (or better another column), to shape the data.
How could I make a cluster of "actions", in order to divide them in 4 days?
And after I have reshaped, I should use the xtset comand like
xtset people days or xtset people days, cluster(actions)?

This would be only the beginning of my regression, but I need and help to start.
Thank you for your attention, I hope that someone will be able to give me an advice. In case you need any other explanation, feel free to ask of course