probit panel data

[adelia wahu hernawati]

Carlo Lazzaro, i want to ask how to estimation probit panel data. based on the data I have I use panel data probit estimation. therefore I use the command "xtprobit"
but there are still questions that block me, with the independent variables I use UMR, total labor, taxes and other expenses (GCD).
In the panel data that I have for variables such as age, taxes, and other expenses, there are still many gaps in the survey data (based on survey data from large and small industries) that I have, so it is assumed to be '0' when I directly estimate with probit panel data, so results are not significant. however, when I log the independent variables, the probability results are significant, but the number of observations and also the groups are different from those that I did not log.

does the value of 0 have too much influence on the estimate madam? Is it permissible to log variables, even if they affect the number of observations?
here I permit to send a comparison of estimates without logs and with logs along with the sum of the variables.

[Carlo Lazzaro]

Adelia:
both addessing a thread personally and pasting screenshots are deprecated behaviours on this forum.
Please see the FAQ on how to post more effectively.
Tha said:
1) are you dealing with a panel or a survey dataset?
2) replacing missing values with zero is an arbitrary and often dangerous approach. You should investigate why values are missing, what is the mechanism (and the pattern) underlying their missingness, and whether (or not) their missingness is informative. Briefly, is much better to leave data as they are (if missing, Stata will get rid of the entire observation, reducing your sample size accordingly) instead of making them up;
3) as the natural log of zero does not exist, logging reduces your sample size even more;
4) striving for statistical significance is not scientific; results are what they are and statistically non-significant coefficients are as informative as their significant counterparts;
5) Carlo (he/him).