Page 1 of 1

Input columns have unequal lengths

Posted: Tue Jul 14, 2020 9:34 am
by DanielHolman

I just updated both my machines to the latest version of mlwin, which I have installed so that I can use runmlwin from stata. On one machine, I have 3.01 installed, and on the other it was an older version, 2.x (can't remember which exactly).

Now I am getting this error with my models, where I didn't previously. I wonder whether this is due to the following being listed as an update from 3.03 onwards: 'Check that all column lengths are equal when running the GMUL command' (though I'm not exactly sure what GMUL is).

Is it possible this is a bug with the programme? If so, can I revert back to a previous mlwin version? The old folders seem to have disappeared since I have upgraded. Obviously I don't want to do this if the models based on the old version were invalid, but as it stands I'm stuck.

Note I guess I would have the same problem with mlwin standalone, but I'm only familiar with runmlwin.

PS to add this is a 3 level cross-classified model. I don't get the error if I run the same model without it being cross-classified.

Many thanks

Re: Input columns have unequal lengths

Posted: Mon Jul 20, 2020 2:11 pm
by ChrisCharlton
I'm not sure why this is happening. Can you confirm whether the examples provided with runmlwin work correctly (i.e. those under help runmlwin and on ... /examples/? If they do could you provide examples of the syntax this is not working correctly for you? Can you also confirm that your version of runmlwin is up to date?

Re: Input columns have unequal lengths

Posted: Fri Jul 31, 2020 1:11 pm
by DanielHolman
Hello just to update on this in case useful for others, it was caused by the number of missing cases for the level 1 and level 2 variables not matching up. Once I dropped missing cases to use a complete case dataset, the problem disappeared. Thanks anyway for the reply. Dan.

Re: Input columns have unequal lengths

Posted: Mon Aug 03, 2020 10:13 am
by ChrisCharlton
Thanks for the update. This is a bit strange as, unless the model is multivariate, MLwiN should be essentially listwise deleting the rows with missing data anyway. If at some point you could provide me an example that exhibits this problem I will investigate further.