HELP! Is using cross section semi series data correct?

#1
Hello all! It is so nice to know that there is such a forum to discuss statistical matters.

Allow me to introduce myself. My Name is Yohanes (John) from Indonesia. So, Greetings from the South East Asia. And Yes, I'm not a very good English speaker so please spare my grammatical errors.

I recently doing a study in the stock market regarding repurchases and I stuck for almost 3 months. I am hoping any of you could give me suggestions or even solution to this problem.

My research is about the liqudity of the stock market when the repurchase happen. However, since repurchases only happen 5% of the time in, so it will be almost impossible for it to affect the whole market.

Moreover, the repurchases happen in a program (18 months) and there will be a transaction everyday, for a single company.

How I do the research:
1. I separate transactions between non-repurchase days and repurchase days, since it is impossible to compare a trend of 5% population to the whole.
2. I regressed the whole thing, and the result is quite significant.

However, my question is this:
1. Can I really just prove that there is a difference in the nature of transaction between non repo (95%) and repo days (5%)only by comparing its means? Is this legit enough?

2. Within this 5% of repurchase data, it is a cross sectional, semi series data. So let's say that the repurchase transaction only happens in the 1st, 2nd, and 5th day of transaction. Is there any issue for using a data like this?

Thanks before!
 
#2
What do you mean by "the nature of transaction"? An expected return, a return premium, volume, ...? You should pool stocks engaged in repurchases and stocks not engaged in repurchases in one data set. You should follow those stocks over time. For that reason the data would be a panel. So you would have to use methods for panel data. If you do, you can explore potential abnormality. Happens all the time in the empirical finance literature.
 
#3
What do you mean by "the nature of transaction"? An expected return, a return premium, volume, ...? You should pool stocks engaged in repurchases and stocks not engaged in repurchases in one data set. You should follow those stocks over time. For that reason the data would be a panel. So you would have to use methods for panel data. If you do, you can explore potential abnormality. Happens all the time in the empirical finance literature.
Thanks for your answer!
Youre right. What I meant by nature of transaction is the abnormalities if theres a repurchase present.
I will try to use panel data to proceed my research. Do you happen to have any book or reading reference of exploring potential abnormalities in panel data?
Or can I just pool all data just like you said and set a dummy: 0 for no repurchase and 1 for repurchase?