Making groups and comparing groups in Stata

[Dirk Harmelink]

Hello everyone,
I hope this forum can help me, i've seen some helpful answer so I thought I might try as well.

I have a panel data set (N = 520, T = 983), where each firm has its own "lottery-like-score" per day. I want to test with covid dummies whether trading volume (in times when traditional gambling opportunities are limited, due to COVID for example) for stocks scoring high on lottery-like-score increases significantly more than for stocks that have a lower score.
I figured I design three groups and average the trading volume for each group, meaning that I end up with a panel dataset of N = 3 (high LL, other stocks, Low LL) and T = 983.
Is there a way to desing those groups?

Thanks in advance!

[George Ford]

It's always iffy to dichotomize a continuous variable. There are always arbitrary cutoffs involved. I suspect the firms will move in and out of groups as well.

A finite mixture model may be useful, at least in the sense it gets you out of it.

Or, you can cycle through a bunch of alternatives and see what gets you the highest R2. This is proposed in Pindyck and Rubinfeld (3rd) at p. 120.

[George Ford]

egen cut would get your groups.