Long term, this data structure may prove
awkward and some reshape (or -reshape-)
may be worthwhile.
With your existing structure, and if
I understand this correctly:
1. Take the variable list and extract
substr("<varname>",3,4)) in each case.
foreach v of var * {
local mid = substr("`v'",3,4)
local mids "`mids' `mid'"
}
2. That list will have repeated values,
so we slim it down:
local mids : list uniq mids
and sort it at the same time
local mids : list sort mids
(see help on -macrolists-)
3. Now we cycle through those
middle bits:
qui foreach m of local mids {
unab this : ??`m'??
local nthis : word count `this'
if `nthis' {
gen m`m' = 0
foreach v of local this {
su `v', meanonly
replace m`m' = m`m' + r(mean)
}
replace m`m' = m`m' / `nthis'
}
}
(not tested!)
Dietrich Dr�ner
> my data set contains 113 variables that belong to 62 cases.
> There are 1 to 5
> variables per case.
> I need to calculate the mean of the means of these 1 to 5
> variables for each
> of my 62 cases.
>
> -----------------------------------------
> A possible code for one case with 3 variables looks like this:
>
> sum sd0032i1
> local x=r(mean)
> sum sd0032i3
> local y=r(mean)
> sum sd0032i4
> local z=r(mean)
>
> gen msd0032=(`x'+`y'+`z')/3
>
> --------------------------------------
>
> The 3rd to 6th letter of the variable names is identical for
> the variables
> of each case, the last letter distinguishes the 1 to 5
> variables with an
> ascending order but unfortunately sometimes with "gaps" like
> in the case
> above.
>
> My problem is: how can I generalise this - or an equivalent -
> code to be
> repeated 62 times with a varying number of variables?
>
>
*
* 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/
