[William Lisowski]

Yes, I misremembered the rc that Chris got.

What is interesting is that the "invalid syntax" error message is not consistent with rc 111 - I thought it was some sort of generic, but in fact it is the precise message for rc 197 - reporting that there was a syntax error on the syntax command (or some other parsing command). And in my Stata 16.1 installation, a missing which returns rc 111.

. capture noisily which gnxl
command gnxl not found as either built-in or ado-file

. display _rc
111

And for you too it printed the "invalid syntax" message associated with rc 197 and then reported rc 111 on exit.

help rc 111
help rc 197

Let me add that I have the same experience with the 1.5.1 version of matchit, so it is a pre-existing condition.

. capture noisily which matchit
/Users/lisowskiw/Library/Application Support/Stata/ado/plus/m/matchit.ado
*! 1.5.1 J.D. Raffo April 2019

. capture noisily which freqindex
command freqindex not found as either built-in or ado-file

. matchit idmaster cnms using `myusing', idusing(idusing) txtusing(cnms) di sim(token)
freqindex not found.
matchit requires freqindex to be installed. You can get it in SSC.
invalid syntax
r(111);

I am beginning to wonder if there is a problem in the built-in which command, or else with the capture prefix. As I initially experienced, the two message about freqindex were suppressed in my output. Yet now I too get those messages. And my first attempt just at the code immediately above suppressed the output from the which matchit command. But a second attempt a few minutes later produced what's above.

This may be a problem worth seeking advice about from Stata Technical Services.

Meanwhile, if Chris Jiao does not confirm my hypothesis about freqindex being missing, we still have to discover the source of the original problem. But I'm optimistic that the two are somehow related.

[Chris Jiao]

Thanks William Lisowski and Julio Raffo for your updates.

I double check with my code. Here is what I find.

1. The error r(197) can be resolved by restarting Stata.

2. I have installed both matchit and freqindex via code

ssc install freqindex

and

ssc install matchit

. However, when I run diagnose, I still cannot have the analyzing result (list of most frequent grams in master and using file). But actually, as I said previously, I can run the following code to have frequent grams in another way.

local myN=_N
preserve
freqindex cnms, sim(ngram, 3)
gen share=freq/`myN'
gsort -freq
list in 1/20
restore

So I suppose it is not the problem of not installing freqindex. I speculate that it is one specific problem in my device?

[Julio Raffo]

Dear both,

Chris Jiao I think matchit is not yet updated in SSC (I've sent to them only yesterday). so that's why you're still getting the same results. If you cannot wait, you can install manually or from from Github by using the command github (by E. F. Haghish) that is in SSC:

github install julioraffo/matchit
which matchit

Check that version is 1.5.2 from May 2020, after using the which command. This should fix the issue.

William Lisowski, thanks a lot for the debugging assistance. Clearly I'm doing something wrong with the interaction between capture and error commands. And, yes, I'm pretty sure this bug exists since I put that condition a long time ago. However, it is an innocuous bug as matchit is supposed to stop if freqindex is not found and the messages seem to display (at least in Stata 12, 13, 14 & 15, not sure in 16). The fact that it shows the wrong code is not great and I'll fix it for a following patch.

[Julio Raffo]

Dear both,

Chris Jiao I think matchit is not yet updated in SSC (I've sent to them only yesterday). so that's why you're still getting the same results. If you cannot wait, you can install manually or from from Github by using the command github (by E. F. Haghish) that is in SSC:

github install julioraffo/matchit
which matchit

Check that version is 1.5.2 from May 2020, after using the which command. This should fix the issue.

William Lisowski, thanks a lot for the debugging assistance. Clearly I'm doing something wrong with the interaction between capture and error commands. And, yes, I'm pretty sure this bug exists since I put that condition a long time ago. However, it is an innocuous bug as matchit is supposed to stop if freqindex is not found and the messages seem to display (at least in Stata 12, 13, 14 & 15, not sure in 16). The fact that it shows the wrong code is not great and I'll fix it for a following patch.

small correction, it seems that the github command is not in SSC (apologies for the mistake), but actually at Github: https://github.com/haghish/github#2-...-github-module. You can install it directly from Stata by running:

net install github, from("https://haghish.github.io/github/")

[William Lisowski]

First, my thanks for the pointer to the github command.

I can now confirm that after uninstalling matchit and freqindex, then relaunching Stata, then installing matchit from github, the error message now was consistent with the return code.

. matchit idmaster cnms using `myusing', idusing(idusing) txtusing(cnms) di sim(token)
command freqindex not found.
matchit requires freqindex to be installed. You can get it in SSC or Github.
r(162);

And after installing freqindex from github, the sample code from post #13 ran as expected, with complete output.

The versions of the two commands at this moment are

. which freqindex
/Users/lisowskiw/Library/Application Support/Stata/ado/plus/f/freqindex.ado
*! 1.3.2 J.D. Raffo April 2019

. which matchit
/Users/lisowskiw/Library/Application Support/Stata/ado/plus/m/matchit.ado
*! 1.5.3 J.D. Raffo May 2020

[Awais Ahmed]

Hi Julio, @Julio Raffo & Chris Jiao
I have issue in the below code for some reason it doesn't find id1?? Do you have any advice? Thank you for help.

************************************************** ******************************
use PPP_dataset, clear
sort businessname
duplicates drop businessname, force
g id1 = _n
save PPP_dataset_clean.dta, replace
************************************************** ********************************
use Lobbyview_dataset_Client_Name.dta
rename Client_Name businessname
sort businessname
*drop if businessname == businessname [_n-1
g id2 = _n
keep businessname id2

**reclink
ssc install reclink
reclink businessname using PPP_dataset_clean.dta, gen(myscore) idm(id1) idusing(id2)

[William Lisowski]

The output of help reclink tells us about the idmaster() and idusing() options:

    idmaster(varname) is required and specifies the name of a variable in the master
    dataset that uniquely identifies the observations.  This variable is used to track
    observations.  If a unique identifer does not exist, one can be created simply as
    gen idmaster=_n.

    idusing(varname) is required and specifies the name of a variable in the using
    dataset that uniquely identifies the observations analogous to idmaster.

Since your code has

reclink businessname using PPP_dataset_clean.dta, gen(myscore) idm(id1) idusing(id2)

idusing must be the id variable in PPP_dataset_clean.dta - id1- and idm must be id variable in the dataset in memory - id2 - so your reclink command should be

reclink businessname using PPP_dataset_clean.dta, gen(myscore) idm(id2) idusing(id1)