Forval Loops Breaks because of irregularly spaced values

[Attaullah Shah]

I want to execute the following loop

forval i = 1/500 {
qui sum Ri if PF`i' > 10
loc LF = r(mean)
...
...
...
}

The problem arises when PF`i' is not regularly spaced. How can I force the loop to continue and ignore if PF`i' has missing sequential number i.e. if PF1 and PF2 are followed by PF4, the loop will stop.

[daniel klein]

You need to be more specific about what exactly you want to do if PF<i> is missing, but here is some technique

forv i = 1/500 {
cap conf v PF`i'
if (_rc)  continue
qui sum Ri if PF`i' > 10
...
...
...
}

Best
Daniel

[Svend Juul]

If the PF* variables are consecutive in the dataset, you can use -foreach- instead. If needed, you can make them consecutive with -order-:

order PF* , sequential
foreach i of varlist PF1-PF500 {
qui sum Ri if `i' > 10
loc LF = r(mean)
...
}

[Attaullah Shah]

Daniel you are my hero. Thank you Daniel and Svend Juul for your excellent replies. I made a mistake in my post. Actually, I wanted to ask if a variable is not sequentially placed, then how the forval loop will continue. So my request should read

forval i = 1/500 {
qui sum Ri if PF==`i'
loc LF = r(mean)
...
...
...
}

[Nick Cox]

On a different note:

If all you want after summarize is the mean, use the meanonly option. Even if you want some other stuff, still consider the meanonly option.

See e.g. http://www.stata-journal.com/sjpdf.h...iclenum=st0135

[Svend Juul]

If the problem isn't absence of variables in a sequence, but the ordering of variables numbered sequentally, there is no problem. PF427 is PF427 no matter where in the list of variables it is located.

To Nick: I avoid telling (when teaching) other people about the meanonly option; its name is confusing, and quietly can be used generally. meanonly saves computer, time, right, but nowadays brain time is much more costly than computer time.

[Abraham Wolde-Tsadick]

Hi Attaullah,

You can use -levelsof- to capture distinct values of the variable PF and then run the loop using -foreach- instead of -forvalue- as follows:

levelsof PF, local(PFlevels)
foreach i of local PFlevels {
    qui sum Ri if PF==`i', meanonly
    loc LF = r(mean)
    ...
    ...
    ...
}

[Nick Cox]

Svend: I take your point. Attaullah has elsewhere made it evident that he is interested in learning to program in Stata, and many others are too.

In the 2007 paper cited I suggested a different name.

[Svend Juul]

Attuallah: Re-reading your post #4: You write about the sequential placement of variables, but in the code you have a single variable, PF, which can take different values.

[Attaullah Shah]

Thanks for your replies.
To Abraham: I have used levelsof previously, but it considerably slows things down.
To Nick : Yes I do use meanonly option, in the example I just wanted to fill space with some stata command to complete the example of loop
To Svend : You are right, I did not clarify my point enough, I am interested in using distinct values of the variable PF, not different variables.

So my problem remains somehow unresolved.

[Svend Juul]

Attuallah: Didn't Abraham's suggestion in post #7 do what you want?