changing string to float

[Alauddin Majumder]

Hi all, greetings! I have a variable named "pcgdp" stored as string. Observations are numeric values. Some observations are missing and indicated by double dot "..". I want to convert it to numeric form. Executing the command "destring pcgdp, replace" returns the following: "pcgdp contains nonnumeric characters; no replace". Any help will be greatly appreciated.

[William Lisowski]

I would suggest the following approach.

Start by running

tabulate pcgdp if missing(real(pcgdp))

to confirm that the only non-numeric values in your data are the ".." entries. Once you're assured that there isn't something else in your data that you need to account for, then

destring pcgdp, replace force

will do what you want and will create Stata numeric missing values for the ".." entries.

[Dirk Enzmann]

Another possibility that does not require you to check the results of tabulating the values of pcgdp would be

replace pcgdp if pcgdp==".."
destring pcgdp, replace

The replace command will show you how many occurences of ".." have been replaced and destring will run without error if all other values of pcgdp have been successfully converted into numeric (the number of missing values generated should match the number of ".." replaced into "").

[William Lisowski]

The code in post #3 was intended to be

replace pcgdp = "" if pcgdp==".."
destring pcgdp, replace

Speaking for myself, in the face of a dataset that mixes numeric and character representation for a single variable, I'm reluctant to assume there are not other potential issues, and hence the code in post #2 is my favored approach.

[Dirk Enzmann]

William Lisowski : Of course, I missed = "", thanks for noting and correcting.

[Alauddin Majumder]

Thanks you so much, Lisowski and Enzmann, for useful suggestions.