Means per multiple variables and multiple observations, with missing values

[Mink Sjamsoedin]

Dear Statalist,

I am very new to STATA and Statalist, so please forgive me if in some way my question is unclear.


I am using STATA17 and want to calculate the means, standard deviations and p-values of a dataset combining multiple variables and multiple observations with missing values in between.

I have conducted a survey, for which participants were asked to rate achievement, on a 5-item Likert scale with "0" as the option for "not applicable", for 47 criteria (my variables). I have recoded for "0" to be missing values and now I would like to calculate means for subgroups, combining the multiple variables. I have familiarized myself with

mean()

and

rmean()

and

rowtotal()

But these will only allow for either the rows or columns. How can I get a mean (with weighted missing values) for both rows and columns combined in STATA? I can do it by hand (or calculator) but would like to get the additional information on standard deviation and p-value.

Thank you in advance for your advise.

Kind regards,
Mink Sjamsoedin

[Nick Cox]

You could copy the data to Mata and there take a mean over a matrix. Otherwise, a mean over observations and over variables needs to be calculated in two steps, say by calculating a row mean and then a mean over the row means.

Treating not applicable as 0 is puzzling to me, but it's your data and your problem and your field.

[Mink Sjamsoedin]

Dear Nick,

Thank you for your reply. I am glad to hear that there is no other way than to do it in two steps when using STATA, now I can stop looking for a solution there. I will look into Mata then.
In reply to your remark on 0 for not applicable, this was the only way it could be exported from the online survey service I used, unfortunately.

[Nick Cox]

But now you are in Stata, you can replace 0s by missings. See

help mvdecode

[Mink Sjamsoedin]

Sorry, I should have mentioned specifically that I have already changed the 0 into missing values