Splitting Variables Into Groups

[Bryan Shelly]

I have a list of covariates " v* " all of which need to included in IRT models. In some of the data sets I am using v* contains over 2,000 variables, so I need to divide v* into smaller groups of ~100 variables per group. Given the amount of data I'm working with, "keep" and "drop" are not practical options if they involve naming specific variables in v* or even variable ranges with specific starting and ending variables. I tried the code below from http://www.stata.com/support/faqs/da...ts-of-dataset/ but got a "Group not found" r 111 error.

preserve forval i = 1/7 { keep if group == `i' save group`i' restore, preserve }
Thanks in advance for any help, and apologies in advance for bringing up a topic I know must have been covered many times here. I could not find the right search term to yield the thread that would help.

Bryan

[Rich Goldstein]

we can't tell what you actually typed (as shown above it is illegal); please read the FAQ on how to show code within CODE blocks and try again

[Bryan Shelly]

Apologies. Proper format should be below.

I have a list of covariates v* all of which need to included in IRT models. In some of the data sets I am using v* contains over 2,000 variables, so I need to divide v* into smaller groups of ~100 variables per group. Given the amount of data I'm working with, "keep" and "drop" are not practical options if they involve naming specific variables in v* or even variable ranges with specific starting and ending variables. I tried the code below from http://www.stata.com/support/faqs/da...ts-of-dataset/ but got a "Group not found" r 111 error.

  preserve 
  forval i = 1/7 {
          keep if group == `i'
          save group`i'
          restore, preserve 
  }

[Rich Goldstein]

it appears that there is no variable called "group" in your data - you need to set one up; you don't give enough information about your variable names for us to know whether something simple will do

[Clyde Schechter]

There may be a simpler way to do this, but the following should work:

local group_size 100

unab varlist: _all

tokenize `varlist'
local counter 1
while `"`1'"' != "" {
    local group
    forvalues i = 1/`group_size' {
        local group `group' `1'
        macro shift
    }
    preserve
    keep `group'
    save `roup`counter'
    des
    restore
    local ++counter
}

[Bryan Shelly]

Variables names are "id" "gender" "race" and the variables of v*. The variables of v* are all named using the letter v and a number, so "v70" "v83" etc. The numbers after the v are neither consecutive nor spaced at regular intervals.

[Bryan Shelly]

Clyde, thank you. Stata returned the following:

invalid 'roup1'
r(198);

Thoughts?

Thanks,

Bryan

There may be a simpler way to do this, but the following should work:

local group_size 100

unab varlist: _all

tokenize `varlist'
local counter 1
while `"`1'"' != "" {
local group
forvalues i = 1/`group_size' {
local group `group' `1'
macro shift
}
preserve
keep `group'
save `roup`counter'
des
restore
local ++counter
}

[Clyde Schechter]

Drat! There's a typo. The line that reads -save `roup`counter'- should read -save group`counter'- Sorry about that.