The model did not run properly in MLwiN. You most likely clicked the 'Abort Macro' button in MLwiN, rather than the 'Res

[annajoseph]

Hi,

I used the command "runmlwin (suboanw_nse seom_dum baesbuso_nse msseom_cy mbaesbnse_cy female mfemale_cy educ_secpost meduc_secp_cy educ_post meduc_post_cy age_clean mage_cy age_squared magesq_cy gemwork_dum mgemwork_cy busangyy mbusang_cy knowent_dum mknowent_cy l_xconst l_gov_spend lag_ln_gdppc lag_dgdp cons, equation(1)) (sestup_dum seom_dum baesbuso_nse msseom_cy mbaesbnse_cy female mfemale_cy educ_secpost meduc_secp_cy educ_post meduc_post_cy age_clean mage_cy age_squared magesq_cy gemwork_dum mgemwork_cy busangyy mbusang_cy knowent_dum mknowent_cy l_xconst l_gov_spend lag_ln_gdppc lag_dgdp cons, equation(2)), level1(i:) level2(country_year: cons) discrete(distribution(binomial binomial) link(logit) denominator(cons cons))
"

STATA has ran for more than 24 hours and told me that "The model did not run properly in MLwiN. You most likely clicked the 'Abort Macro' button in MLwiN, rather than the 'Resume Macro' button"

But I did not clicked any button in MLwiN actually. What is going wrong here?

Many thanks!!!

[ChrisCharlton]

Your command does not have the nopause option, which means that the modelling fitting should pause at various points in the MLwiN interface. Were any error messages displayed, and did the model equation look correct?

Have you tried simplifying the model (for example removing predictors) to confirm that this is able to fit?

[annajoseph]

Thanks for your reply!

But...

I tried to use a small sample of my data to run. After I did the command runmlwin, the program MLwiN showed up. And I clicked the button resume macro, some coefficients came out in the MLwiN. But then the program MLwiN became frozen, showing "not responding"......

And I just tried nopause using the small sample. MLwiN froze again..."not responding"

[annajoseph]

And.....

When I used the full data, and nopause command, MLwiN said " overflow allocating smatrix"

What should I do?

Many thanks!!!

[ChrisCharlton]

It may be that MLwiN is still working and just not yet updated the interface. I would suggest leaving it for longer to see whether it becomes responsive again, as well as checking the Task Manager to see whether it has any CPU activity.

[annajoseph]

Hi Chris,

Thanks for the advice!

I checked the task manager which also displayed "not responding"

And I just left it running, after an hour it just gave me an error message and no result...

What should I do?

Thank you so much!

[ChrisCharlton]

Could you provide details regarding what the error message said?

I would also suggest testing some of the examples from http://www.bristol.ac.uk/cmm/software/r ... /examples/, as this could help determine whether the problem is specific to your machine or something to do with your data/model.

[annajoseph]

Hi Chris!

Thanks for your advice!

I have tried a lot of examples from the site that you gave. I think the machine is ok!

I think this is specific to my data and my commands.

I used my full data (size:around 170,000). when I ran runmlwin (suboanw_nse seom_dum baesbuso_nse msseom_cy mbaesbnse_cy female educ_secpost meduc_secp_cy educ_post meduc_post_cy age_clean mage_cy age_squared magesq_cy gemwork_dum busangyy mbusang_cy knowent_dum l_xconst l_gov_spend lag_ln_gdppc lag_dgdp cons, equation(1)) (sestup_dum seom_dum baesbuso_nse msseom_cy mbaesbnse_cy female educ_secpost meduc_secp_cy educ_post meduc_post_cy age_clean mage_cy age_squared magesq_cy gemwork_dum busangyy mbusang_cy knowent_dum l_xconst l_gov_spend lag_ln_gdppc lag_dgdp cons, equation(2)), level1(i:) level2(country_year: cons) discrete(distribution(binomial binomial) link(logit) denominator(cons cons)) nopause

MLwiN always told me overflow allocating smatrix

And I also tried to run simple binomial response, e.g. runmlwin suboanw_nse mssest_cy baesbuso_nse mbaesbnse_cy seom_dum msseom_cy female mfemale_cy educ_secpost meduc_secp_cy educ_post meduc_post_cy age_clean mage_cy age_squared magesq_cy gemwork_dum mgemwork_cy busangyy mbusang_cy knowent_dum mknowent_cy fearfail_dum mfearfail_cy suskill_dum msuskill_cy l_xconst l_govt_size lag_ln_gdppc lag_dgdp cons, level1(i:) level2(country_year: cons) discrete(distribution(binomial) link(logit) denominator(cons)) nopause

STATA gave me a result.

Could you please give me some suggestions?

Yuan

[ChrisCharlton]

This error usually means that one of the matrices that MLwiN needs to allocate for estimation would be too large. Possible things to try are:

1. Add the optimat option to your -runmlwin- command.
2. If you aren't already doing so switch to using the 64-bit version of MLwiN
3. Examine your data to see whether any of your higher level units are particularity large. If so try using a sub-sample.

[annajoseph]

Thanks Chris!!!!!!

It worked with the the optimat option !!!!!!

Can I ask another question?

The result does not show Log Likelihood and Wald Chi sq.

Do you know how I can get the two statistics?

Thank you very much!

Yuan