### Thursday, March 09, 2006

## st: RE: differentiating missing values in codebook

I can't comment on SPSS, which I believe to be another program.

But there is no need for hunch or surmise.

Even if you haven't got a manual to hand, a good introduction to Stata's approach to missing values is provided by

. help missing

In essence you can map those large negative values to specially coded missing values of your choice, after which you can, I imagine, emulate anything reasonable that your other program produces fairly closely.

Nick n.j.cox@durham.ac.uk

Graham Wright > I'm trying to use stata to produce a weighted codebook of a large > dataset. Just a lot of one way tabs in survey mode. However, > I'm having > trouble getting the missing values to work the way I would like. > The data was originally coded in SPSS, and missing values > were expressed as large negative numbers (-999 etc) and given > different > labels depending on why the data is missing ("Refused," "Question not > asked" etc). Naturally, when I run frequency commands in SPSS I get > relative frequencies for the valid data only, but I get to see the > missing values too. However, since we're using a complex > sample design, > I need to use stata to get the correct confidence intervals, > but as far > as I can tell, the only value that stata will accept as missing is a > period, anything else gets included in the analysis, so I can't > differentiate between different kinds of missing values, like > I can in > SPSS. Is there a way to get an SPSS style output (valid, > missing & valid > percent, total percent) in Stata? My hunch is that you simply > can't do > this in Stata but maybe someone knows a way.

* * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

Tag: statalist