Here is a program that'll do what you want (and more). Save the program in a separate file called conddropvar.ado on your hard disk. In the example change the -cd- command to point to the folder in which you stored conddropvar.ado.
Hope this helps,
Maarten
*----------- begin program -------------
*! 1.0.0 MLB 6 Jul 2007
program define conddropvar, rclass
syntax [varlist] [if] [in] [aweight fweight iweight] , ///
[ ///
ALLOBS(string) ///
MINVAR(numlist>=0 max=1) ///
MINVALUES(numlist>0 integer max=1) ///
STROK ///
IGNORESTR ///
REPLACE ///
]
marksample touse, strok
qui count if `touse'
local N = r(N)
if `N' == 0 {
di as error "no observations"
exit 2000
}
if "`weight'" != "" {
local wgt "[`weight'`exp']"
}
if `"`allobs'`minvar'`minvalues'"' == "" {
di as error "at least one of the options allobs, minvar, and minvalues"
di as error "must be specified"
exit 198
}
if `"`strok'"' != "" & `"`minvar'"' != "" {
di as error "options strok and minvar cannot be combined"
exit 198
}
if "`strok'" != "" & "`ignorestr'" != "" {
di ass error "options strok and ignorestr cannot be combined"
exit 198
}
if "`ignorestr'" != "" {
qui ds `varlist', not(type string)
local varlist "`r(varlist)'"
}
if `"`allobs'"' != "" {
if "`strok'" != "" {
capture confirm string variable `varlist'
if _rc {
di as error "All variables in varlist must be strings"
di as error "if both the strok and allobs options are specified"
exit 198
}
}
if "`strok'`ignorestr'" == "" {
capture confirm numeric variable `varlist'
if _rc {
di as error "No variable in varlist may be a string"
di as error "if the option allobs is specified and"
di as error "neither the options strok or ignorestr are specified"
exit 198
}
}
}
if `"allobs"' != "" {
gettoken var : varlist
local cond : subinstr local allobs "var" "`var'", all
capture sum `var' if `cond' & `touse'
if _rc > 0 {
di as error "`allobs' is not a valid condition"
exit 198
}
}
if `"`minvalues'"' != "" {
tempvar first sort
gen long `sort' = _n
gen byte `first' = 0
}
foreach var of varlist `varlist' {
local drop = 0
if `"`allobs'"' != "" {
local cond : subinstr local allobs "var" "`var'", all
qui count if (`cond') & `touse'
if r(N) == `N' local drop = 1
}
if `"`minvar'"' != "" {
qui sum `var' `wgt' if `touse'
if r(Var) < `minvar' local drop = 1
}
if `"`minvalues'"' != "" {
qui replace `first' = 0
qui bys `var' `touse' : replace `first' = 1 if _n == 1 & `touse'
qui count if `first'
if r(N) < `minvalues' local drop = 1
}
if `drop' == 1 local todrop "`todrop' `var'"
}
local todrop : list retokenize todrop
if "`replace'" == "" {
di as txt "the replace options was not specified so the data remain unchanged"
di
if "`todrop'" != "" {
di as txt "If the replace option was also specified"
di as txt "the following variables would have been dropped: "
di as result "`todrop'" as txt "."
di
di as txt "To drop those variables repeat this command with the replace option."
}
else {
di as txt "No variable would have been dropped"
di as txt "if the replace option was specified."
}
return local todrop "`todrop'"
}
if "`replace'" != "" {
if "`todrop'" != "" {
di as txt "The following variables have been dropped: "
di as result "`todrop'" as txt "."
drop `todrop'
}
else {
di as txt "No variable has been dropped"
}
}
if "`minvalues'" != "" {
sort `sort'
}
end
*------------------ end program -----------------
*---------------- begin example ---------------
set more off
local home "h"
cd "`home':\mijn documenten\projecten\stata\conddropvar"
sysuse auto, clear
*set trace on
conddropvar , allobs("var==0 | missing(var)") ignorestr
*set trace on
conddropvar , allobs("var==0 | var==1") minvalues(6) ignorestr replace
exit
*-------------------- end example -------------------
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
-----Original Message-----
From: [email protected] [mailto:[email protected]]On Behalf Of Guillermo Villa
Sent: vrijdag 6 juli 2007 16:22
To: [email protected]
Subject: st: RE: Drop variables satisfying a condition
Hi,
Thanks, but I think the loop does not work with strings, so it stops when it
finds the first string in the dataset. Do you think it is possible to write
a similar loop for any variable format? I have a huge dataset, with about
500 variables, so this will be very helpful.
Thanks again.
Guillermo
-----Mensaje original-----
De: Guillermo Villa [mailto:[email protected]]
Enviado el: viernes, 06 de julio de 2007 14:33
Para: '[email protected]'
Asunto: Drop variables satisfying a condition
> Dear statalisters,
>
> I wonder whether it is possible to drop several variables in a dataset
> specifying an "if" condition. For instance, it would be interesting
> being able to drop all variables in the dataset for which all cases
> are 0 or missing values. I have been reading the help documentation on
> the "drop" command, however I am unable to find the right answer so
> far. Note that I want to drop variables and not cases.
>
> Thanks in advance.
>
> Guillermo
>
>
*
* 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/
*
* 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/
