Passing weights won't work with parallel prefix

[gvegayon]

Seems to be a bug.

Will post it on github and take a look at it

Best,

George G. Vega Yon +1 (626) 381 8171 https://ggvy.cl

On Tue, Mar 27, 2018 at 2:47 PM, Liao, Tim F tfliao@illinois.edu wrote: Sure:

. parallel setclusters 7 N Clusters: 7 Stata dir: E:\Program Files (x86)\Stata15/StataSE-64.exe

. parallel: iGini incwage [aw=weight], by(SexRace) weights not allowed r(101);

where iGini is an ado that I wrote that runs well when executed alone.

Ideas? From: George Vega Yon [g.vegayon@gmail.com] Sent: Tuesday, March 27, 2018 4:36 PM

To: Liao, Tim F Subject: Re: Stata Parallel

Can you show me an example of the code you are trying to run?

Best

George G. Vega Yon +1 (626) 381 8171 https://ggvy.cl

On Fri, Mar 23, 2018 at 10:45 AM, Liao, Tim F tfliao@illinois.edu wrote: Hi George,

Thanks for your answer regardless.

I've found that a freshly installed version of parallel would run. However, the parallel doesn't allow the use of weight, which I implemented for computation with sampling weight entered as aweight in my ado. Is there a way to get around it?

Cheers,

Tim From: George Vega Yon [g.vegayon@gmail.com] Sent: Friday, March 23, 2018 1:21 AM To: Liao, Tim F Subject: Re: Stata Parallel

Hi Tim,

So sorry for the delay, your email got lost in my inbox.

That error has to do with the fact that the lparallel.mlib file is not in the right place. I recommend you taking a look at the instructions to install parallel at the github website: https://github.com/gvegayon/parallel

HIH

George G. Vega Yon +1 (626) 381 8171 https://ggvy.cl

On Sat, Feb 3, 2018 at 10:02 AM, Liao, Tim F tfliao@illinois.edu wrote: Hi George,

I've been trying to use the Stata parallel ado that you wrote, without success so far. The error message is always the following:

: 3499 parallel_setclusters() not found even though the installation of "parallel" appears to have gone well. Any suggestions? Thanks, Tim

[bquistorff]

A simple example is

sysuse cars
parallel: reg price mpg [aw=weight]

This happens because parallel_do uses syntax and doesn't catch weight or = exp. We should allow both. The syntax command however will not let you catch both, so we'd have to do more manual string parsing.

[gvegayon]

Sounds about right. Perhaps we can work on this after sending the manuscript... I owe you VERY BAD my final comments! Will work on that on the weekend so we can send it asap (will try to re-run the simulations then on an updated version of Stata...)