Dear friends

am running a GMM regression with 2 lags and observation exceeding 100K .. it takes hours for each run .

do the cloud version do the calculation faster ? or any other suggestion ?

I got tired from it

thank you

Dear friends

am running a GMM regression with 2 lags and observation exceeding 100K .. it takes hours for each run .

do the cloud version do the calculation faster ? or any other suggestion ?

I got tired from it

thank you

That depends, is the function you are using parallelized? If so, do you have more cores available in your Posit Cloud account than in your own machine? If you have the same amount of cores available, the local machine would probably be faster since virtualization servers are usually optimized for core count not per-core frequency.

It is quite possible that your model is having trouble converging. Try changing the starting values, `t0`

, or the convergence criterion, `crit`

.

Sorry but that doesn't answer any of my questions, I don't know what R package you are using to know if the function is parallelized or not, I don't know how many cores your specific CPU model have and I don't know if you have a paid Posit Cloud account that allows you to increase the cpu core count.

It is unlikely that more cores will help much. Try setting `traceIter`

to `TRUE`

and examine the results. You may also want to try setting `itermax`

to 2 and see what happens.

You seem to know which specific function the OP is referring to. If the function is not parallelized (or parallelizable) then I agree, more cores will do nothing.

Presumably, it's gmm from the gmm package. The package doesn't mention anything about parallel.

GMM is basically a fancy nonlinear regression. There might be some part deep inside that runs in parallel, but I doubt it. And if there is, it's probably a piece of matrix manipulation where the overhead of running in parallel would dominate the extra core advantage.

(But I'm guessing.)

Unfortunate that Generalized Method of Moments, and Gaussian Mixture Models have the same abbreviation.

1 Like

Oh. I had assumed it was Generalized Method of Moments. If that's not right, ignore everything I've said. (Good call @nirgrahamuk !)

yes indeed its Generalized method of momentum

I had another issue if you are familiar with 2way robust cluster .. I couldn't figure out the correct package for it .

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.