Searching for a string variable among numeric variables and making changes to the same.

[Arnab Dutta]

I have several variables - some numeric and some string - in my dataset. The string variables are so because the missing values are denoted as "NA" among non missing numeric observations. I want to replace the "NA" by "." and then destring the variables. One way to destring the data would be to use the following:

foreach v of varlist v1 v5 v7 v9 .... {
replace `v' = subinstr(`v', "NA", ".", .)
}
destring, replace

But since there are too many variables its cumbersome to type all of them using foreach. Is there an efficient way to do this?

[Nick Cox]

Often asked here. See e.g. http://www.statalist.org/forums/foru...s-in-a-dataset

But you can solve the problem more directly.

 
destring *, replace ignore("NA")

will ignore numeric variables and map "NA" to missing directly and so cut out the unnecessary conversion to "."

Demonstration:

clear 
set obs 42 
gen nOK = 1 
gen sOK = "frog" 
gen sBad = cond(_n < 42, string(_n), "NA") 
destring *, ignore("NA") replace

[Arnab Dutta]

Thanks a lot. That was really useful.