Create a new variable indicating if all variables in the dataset have the missing values

[David Wang]

Hello,

I have been trying to create a new variable called truemiss to see whether or not ALL the variables in the dataset have missing values. There are 45 variables in the dataset. I first assigned zero to truemiss. Then I used loop to go through each variable, and assign zero to truemiss if the variable is missing (indicated by "") ; otherwise, assign 1 to the truemiss. Last, I attempted to add up the values of truemiss. Below are my codes. For an unknown reason, it is not working. I wondered how to solve this problem? I cannot figure it out.

gen truemiss=0

foreach bimiss of varlist var1-var45 {
    
    recode truemiss=0 if (`bimiss'=="")
    recode truemiss=1 if (`bimiss'!="")
    sum(truemiss)
}

Many thanks,
David

[William Lisowski]

Your description of what you want is unclear.

But perhaps the output of

misstable summarize var1-var45, all

would provide the information you seek about missing values in your data.

[Marcos Almeida]

I believe you just need to use - egen - with the - rowmiss (or - rownomiss) option.

[Nick Cox]

There could be many reasons why the code in #1 is not working.

One is that it assumes that all the variables being looped over are strings.

What the code tries to do, I think, is count the number of missing values in each variable.

But what I think you want to do is orthogonal, to count the number of missing values in each observation. In the latter case Marcos Almeida has a solution.

In the former case, missings (Stata Journal) supplies some functionality and has not yet been mentioned in this thread. See https://www.statalist.org/forums/for...aging-missings for an overview.

[David Wang]

Thank you, Marcos! I followed your advice and it worked perfectly! Below is the code.

egen truemiss =rownonmiss(var1-var45), strok

Thank you, William!
Thank you, Nick!

Best,
David