Before you contact us please do the following:
- Install the most recent update to Stata. Type help update in Stata.
- Install the latest version of SPost. For details, click here.
- Check the discussion of the topic in our book.
- Include only letters, numbers, and underscores in your value labels. This shouldn't be a problem, but sometimes it is.
- Check your descriptive statistics to make sure your variable are what you think they are.
- Try the same analysis with another dataset to determine if your problem is data dependent.
- Check our FAQ Page.
If you still need help, please follow these guidelines fully
We would like to help, but 1000s of people use SPost. want to help, The following things make it easier and faster for us to help. Please follow these suggestions carefully and read Bill Gould's great suggestions on "How to successfully ask a question on Statalist."
- Do not send large files or zip files since they could be rejected by the IU system. Ideally, create a Dropbox folder and send an invitation.
- Create a simple do file that generates the problem you are having. Exclude everything that is unrelated to the problem. Inside the file, include your name and contact information.
- Log the output to a text file, not a smcl file (e.g., log using jslong, text replace). Send the log file along with the do-file.
- Start the do-file with the command -spost13which-.
- Do not refer to specific directories. For example, do not include something like use c:\data\project3\sample.dta because it will not run on our machines unless we either edit your file or create your directory structure (which we won't do). The same is true if you use something like log using c:\data\project3\problem, text.
- Include descriptive statistics for the variable used in the commands that cause a problem.
- Include comments to explain what is going on and what the problem is.
- Your do file might look like this:
capture log close
log using scott_long, replace text
// scott long - 2014-09-01 (email@example.com)
// mtable includes missing values in the predictions
use scott_long_data, clear
// descriptive statistic look fine
desc y x1 x2
sum x1 x2
// logit model gives reasonable results
logit y x1 x2
// the change coefficients for x2 are miss
mtable, at(x1=(0 1))