Create scalar depending on a condition (involving scalars)

[Thomas Mitterling]

Hey,

I want to assign the starting value of my loop to a scalar. The starting value depends on two things:
1. Which row has the xth value of a certain variable (I use "sum" in order to get information on the start of the series).
2. Which row belongs to a certain date (here I use "su" in order to get that information).

Whichever of these two rows is "later" in the series I want to use as starting value.

I tried the following ("sstart" contains the information of condition 1, "r(min)" the information of condition 2):

if r(min) > sstart {
scalar drop sstart
scalar sstart = r(min)
}

which leads to "invalid syntax"

and

scalar sstart2 = r(min) if (r(min) > sstart)

which leads to "if not allowed".

I hope my problem is clear. Thanks in advance!

Best,
Thomas

[Clyde Schechter]

I cannot replicate your difficulty in #1:

. sysuse auto, clear
(1978 Automobile Data)

. scalar sstart = -1

. summ price

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
       price |         74    6165.257    2949.496       3291      15906

. if r(min) > sstart {
.         scalar drop sstart
.         scalar sstart = r(min)
. }

. display sstart
3291

works just fine for me.

Now, that said, you could be encountering your problem if your -summ- command does not immediately precede the -if r(min) > sstart-. Remember that the values stored in r() are volatile: they disappear as soon as you run any other command that stores something in r(). So if there is an rclass command somewhere between your -summ- command and your attempt to reference r(min), r(min) will be undefined, and -if rt(min) > sstart= will be interpreted as -if > sstart- which is, indeed, a syntax error.

As for your second problem, the -scalar - command does not allow -if- qualifiers. In fact, it would make no sense for it to do so. -if- qualifiers tell Stata which subset of the observations in memory are to be used in executing the command. But -scalar sstart2 = whatever- doesn't involve the observations from the data set at all, so even were -if- qualifiers allowed, they would have no effect. What you mean here, I believe, is you want Stata to check whether the current value of r(min) > sstart and then run the -scalar sstart2 = r(min)- command only when that is the case. The code for that involves an -if- command, not an -if- qualifier:

if r(min) > sstart {
    scalar sstart2 = r(min)
}

As with your first issue, this code will only work provided the corresponding -summ- command appears immediately before -if r(min) > sstart {- (or at least no other rclass commands come between them.)

Added: As an afterthought, the -scalar drop sstart- command is not needed and accomplishes nothing.

[Thomas Mitterling]

Thanks for your answer Clyde! Interesting.

My code actually consists of several nested loops. Since I'm not able to debug the code I just commented out the new parts in order to find out what caused the error. The if-part follows directly after the -summ- command.

The code looks like this (the loops runs for all countries in the dataset and the current value is `cc'):

sum obsN if id==`cc'

if r(N)>40 {

scalar smin = r(min)

scalar smax = r(max)

scalar sstart = r(min)+40

su obsN if id == `cc' & date == yq(1990, 2), meanonly

if r(min) > sstart {

[INDENT=2]scalar sstart = r(min)[/INDENT]

}

...

Commenting out the if-part leads to the code running without any problem.

edit:
I found the problem. If no observation fulfills the condition of the -summ- command the variable r(min) does not exist, which leads to an error. Thanks a lot!