Is number of observations (151) in panel data regression enough? does it provide reliable estimations?

[Maha Natsheh]

Hello everybody,
Hope this finds you well!

I have data for 81 agricultural cooperatives for 3 years (2021-2023), but there are numerous missing values, I solve the problem by using replacements of independent variables that have not much missing values, but the problem of dependent variable is remaining (with just 151 observations and 92 missing values!), the overall observation of the regression is also 151 with 76 groups, is this enough to give reliable estimations with the chosen random effects (By Hausman and xtoverid)?

If not reliable, how to compensate or deal with the missing values in dependent variable?

Please note that I have around 8 independent variable, does delete some of them and suffice with 5 dependent variable will mitigate the problem - if existing-? even if this will decrease R^2 ?

Thanks in Advance!
Maha Natsheh

[Clyde Schechter]

Well, even if there were only 1 independent variable, I would be very skeptical of results from a data set where the outcome variable is missing in nearly 40% of the observations. The issue of number of independent variables is mostly in regard to avoiding overfitting of noise in the data. But with nearly 40% of the observations having missing outcomes, you have a much more serious problem: it is likely that everything is biased as a result.

Take a look at https://statisticalhorizons.com/wp-c...aterials-1.pdf for an overview of handling missing data.

[Carlo Lazzaro]

Maha:
in your situation, I do not think that anything more than descriptive statistics is defensible, especially if you want to submit a paper about your research to a technical journal.

[Maha Natsheh]

Thanks very much Clyde, your reply was very helpful! I tried the MI method and I guess it works.
I have a question, How could I access to other topics that listed in the beginning of the file you attached, I need further information about some related topics about MI, it seems I need a subscription to access?

Many Thanks!

[Clyde Schechter]

I'm not sure what you're looking for. You might take a look at https://www.google.com/url?sa=t&sour...te5qinDzS-4Bjo. If that's not suitable, you might be best off using a search engine to zero in on your particular needs.