Estimation sample not in memory - invalid syntax

[Luca Giangregorio]

Hi everyone,

I am using the command inequaly to compute the Gini coefficients with standard error estimates and bootstrapping on them. Everything works perfectly for all countries but one.
Using the loop I realize that this country is the last one, coded as 57. Running the code on this single country as

 inequaly equivalized_hi [pw=rg002] if country==57 & wave==8

I got the following message:

estimation sample not in memory
invalid syntax
r(198);

I also try the option e(sample), but the same result is obtained. If I remove the weights it works perfectly, but I do not have any problem with the weight I am using, in the sense that all observations have their own weight and I cannot understand what could be the problem. Does anyone have any suggestion?
Thank you

[Marcos Almeida]

I know you know it pretty well, but, generally speaking, we have to be pretty sure there is a wave equals to 8 for this specific country.

[Luca Giangregorio]

Hi Marcos, thanks for the reply.
I'm not sure I understand what you mean, but in this specific case there's a wave 8 for that country. I cannot figure out why the command is working without weights, if weights should not present any problem (nor any missing).

[Marcos Almeida]

By the way, I didn’t find any command- inequaly - amongst the adofiles. Did you mean inequal7?

[Luca Giangregorio]

Van Kerm, P. (2016). inequaly - Stata module to estimate relative inequality measures, Statistical Software Components S????, Boston College Department of Economics.
Available from http://ideas.repec.org/c/boc/bocode/sXXXXXX.html.

This is the citation from the help page of the command. inequaly allows for the computation of the standard error and confidence intervals of the estimated measure

[William Lisowski]

The appearance of ???? and XXXXXX where digits would be expected suggests that you are using a version of the software that was not yet generally released. However, searching repec.org does not find a released version, and a quick review of the author's numerous contributions on repec.org does not find anything that looks relevant to my untrained eyes.

https://ideas.repec.org/e/pva19.html#software

So you are it seems working with unreleased user-written software, with which few Statalist members are likely to be familiar. If the help page for the command includes the an email address for support, perhaps you could contact him directly. If not, there's an email address given on his Personal Details page at repec.org linked to at the top of the previous URL.

I note that frequent Statalist contributor Stephen Jenkins has collaborated with Van Kerm frequently; perhaps he will see this and be able to shed further light.

[FernandoRios]

Hi Luca,
Im am aware of the inequaly command. But, I was never able to get my hands on that command. Perhaps you can try the version that I wrote :

ssc install rif
rifhdreg equivalized_hi [pw=rg002] if country==57 & wave==8, rif(gini)

This should give you the Gini and the standard errors associated with it.
Best
It can also estimate other inequality indices and provide standard errors using RIFs

You could also use it to obtain and compare Ginis across waves or countries:

 rifhdreg equivalized_hi ibn.wave [pw=rg002] if country==57, rif(gini) over(wave) nocons

[Luca Giangregorio]

Hi Fernando,
I know about rifhdreg as I use it as unconditional quantile regressions to quantify the impact of each kind welfare regimes on Gini in a pooled analysis and I did not think about it to check it simply the computation of Gini.
Indeed, it works perfectly. Thank you for the hint!

[FernandoRios]

Your welcome!
I made that command for multiple purposes!
Best regards

[Stephen Jenkins]

I am unsure whether Philippe Van Kerm has released inequaly publicly; I'd advise emailing him personally for information. (He doesn't read Statalist.) Another programme that will estimate the Gini coefficient and its SE, and account for sample design features is svylorenz on SSC

[Luca Giangregorio]

Dear Stephen,
thank you for your advise about svylorenz, which I did not think about as well rifhdreg.
You're probably agree on the fact that he does not publicly released it.
Best regards