command hlu

[Dennis]

Hello,
i want to make a test of serial auto-correlation and i know i have to use the command “hlu”.
So I want to transform my equation like
Y(t)’=Y(t) - Y(t-1).
My problem is whether i have to lag my dependent variable separately or if the command “hlu” lags it automatically.
So, is the command:
1. hlu Y X Y(t-1), nolog
or
2. hlu Y X, nolog

[Rich Goldstein]

typing -h hlu- gives the following result:

Out-of-date command

    As of Stata 6.0, the corc, hlu, and regdw commands are out of date.

    The replacement for corc and hlu is prais.

    The replacement for regdw is dwstat after regress.

If this does not help, I suggest reading the FAQ and then re-posting a different question

[Mike Lacy]

What version of Stata are you using? -help hlu- shows that this command is no longer part of official Stata, having been subsumed by -prais- in version 6, which goes back perhaps a decade or so. I do not think your belief that you have to use -hlu- is correct, so knowing why you think this might be helpful. If all you want is the test, try -search durbin watson-.

Regards, Mike

[Dennis]

Okay, first, thanks a lot, but I think durbin-watson doesn't solve my problem.

I'm using Stata 13.1.

For my estimations I have short-term elasticites and I have to transform them in long-term elasticities. Therefore I imitate a procedure which is closely to my examination. In this procedure, the authors use a Hildreth-Lu estimation, so I also want to do so.
What I need is the coefficient ρ to transform the elasticities.

Best regards, Dennis

[daniel klein]

Dennis, type

viewsource hlu.ado

which will show the source code for hlu. As far as I can see the code transforms variables, but it does not add any predictors to the ones specified. So I guess the answer to your question is, you will have to feed your lagged response to hlu.

Whether or not it is wise to use hlu I cannot tell, as the documentation for prais reads

prais fits a linear regression of depvar on indepvars that is corrected for first-order serially correlated
residuals by using [...] a version of the search method suggested by Hildreth andLu (1960).

Best
Daniel