Modelling proportions

[rdmcdowell]

I'm trying to use the online documentation to help with modelling proportions without much success, and wondered if anyone can advice or point me in the direction of other resources.

The outcome is the column which lists the probability of success for each line of data, between 0 and 1
The model selected is binomial, logistic
The 'n' variable is the total number of trials for each line of data eg 500
There are two predictors, the constant, generated as a vector and another predictor.

When I try to run this I get the message: Error while obtaining batch file c:\Program Files\Mlwin\discrete\pre at line number 11....duplicate column name.

However when I look at the column names there are no duplicates and I had no data in the columns it refers to. Can anyone advise? Thanks

[ChrisCharlton]

As part of fitting discrete outcome models MLwiN uses and names a number of columns. You can see an example of this on the line referred to in the error message that you describe:

Code: Select all

name c1180 'H' C1181 'F~(H)' C1182 'H*' C1183 'P' c1184 'MASK' c1189 'Y-VAR' 

If when it gets to this point you already have a column with one of these names in your data then it won't be able to create it again and will display the message that you saw. It is also possible that columns with these names are left over from a previous model run, in which case I would suggest that you remove them prior to fitting the model.

[rdmcdowell]

Thanks for the response. I've noticed that when I start MLWIn for the first time attempting to fit the model, load the data, and check the columns above are indeed empty of data and unlabelled, I still get the same error message.

[ChrisCharlton]

If you can send me an example worksheet, along with instructions on how to replicate the problem I will investigate what is causing this error.