Please use .png for graphic attachments as explained in FAQ Advice #12. Although people who answer Statalist questions can be assumed to have access to Stata, it is harder work to flip back and forth between a graph opened in Stata and this forum. That becomes impossible work if people's version of Stata means that they can't read graphs produced by your version.
You seem unclear about what data you have but there is a limit to which we can tell you anything you don't know or can't expect to find out from the data source(s), which we don't know either.
Jeff Wooldridge guessed, as I think just about any statistical person would do if they focused on the detail, that your Z score was what very often goes under that name, (some variable MINUS its mean) / its SD. Such a z score of necessity would be zero if a value were equal to the mean, and positive and negative respectively for values above and below the mean. That kind of z score is intrinsically not suitable for logarithms.
What you are calling a z score has a very odd distribution with a discontinuity at about 1.6 and a range from about 0.0025 to about 5.8. You should be able to refine these guesses by showing summarize results.
Although on the face of it an entirely positive variable that is positively skew might seem an obvious candidate for log transformation, in your own particular case, the transformation will be likely to produce a tail of outliers for very small values that could produce bizarre side-effects. You can examine this just by a scatter plot of ln z score versus z score and quantile plots.
If I were reviewing this work I would expect an coherent explanation of the nature of this variable before there was any point in proceeding to DID. .
You seem unclear about what data you have but there is a limit to which we can tell you anything you don't know or can't expect to find out from the data source(s), which we don't know either.
Jeff Wooldridge guessed, as I think just about any statistical person would do if they focused on the detail, that your Z score was what very often goes under that name, (some variable MINUS its mean) / its SD. Such a z score of necessity would be zero if a value were equal to the mean, and positive and negative respectively for values above and below the mean. That kind of z score is intrinsically not suitable for logarithms.
What you are calling a z score has a very odd distribution with a discontinuity at about 1.6 and a range from about 0.0025 to about 5.8. You should be able to refine these guesses by showing summarize results.
Although on the face of it an entirely positive variable that is positively skew might seem an obvious candidate for log transformation, in your own particular case, the transformation will be likely to produce a tail of outliers for very small values that could produce bizarre side-effects. You can examine this just by a scatter plot of ln z score versus z score and quantile plots.
If I were reviewing this work I would expect an coherent explanation of the nature of this variable before there was any point in proceeding to DID. .
Comment