Droping observations based on multiple conditions

[Christina Chara]

Hi,

I am trying to correct for small cell size in my data set while adding some exceptions. More specifically, I have small cells in a variable I am interested in and therefore I need to drop those cells that have less than 20 observations in them. I do this using the below code:

gen one=1
egen cellsize=count(one), by (variableofinterest)
drop if cellsize<20

However, I want to keep a selection of cells that have less than 20 observations so I want to tell Stata to drop those with less than 20 observations except the ones I specify. I tried to do this using the below code but it does not do what I want as it drops everything in my dataset

drop if cellsize<20 &variableofinterest!=5 |variableofinterest!=35|variableofinterest!=84

Any suggestions on how I can make this work?

thanks in advance

[Joseph Coveney]

You might want to control the evaluation of your Boolean expressions by using parentheses. Try something like the following and see whether it does what you want to do.

drop if cellsize < 20 & (variableofinterest != 5 | variableofinterest != 35 | variableofinterest != 84)

or

drop if cellsize < 20 & !inlist(variableofinterest, 5, 35, 84)

[Christina Chara]

The second code works perfectly for what I wanted to do.

Thank you very much

[Nick Cox]

Note that

bysort variableofinterest : gen cellsize = _N

is a direct alternative to

gen one=1
egen cellsize=count(one), by(variableofinterest)

Here's yet another way to do it:

egen cellsize=count(1), by(variableofinterest)