using both autofit and difficult in gologit2

[Monika Zukowska]

I would like to know if it is possible to use both autofit and difficult options in gologit2. So far I have only been able to do either one or the other.

An alternative would be to use pl and npl options when setting up my model, and then difficult (In this case I don't need to use autofit anymore), but I haven't found a way to use pl or npl without changing the order of my independent variables.
(i.e. I can use pl once and stuff all the variables that fit the parallel lines assumption in the brackets, but that would make the input order of the variables incorrect. If I use pl multiple times, to preserve the order of variable input, I get and error message)

I am using a do file to run all my analyses. I am new to stata, so its highly possible that I am unaware of some commands/syntax that could help this.
Any help is very welcome, thank you!

[Richard Williams]

Other than aesthetics, the input order of the variables should not matter, i.e. you'll get the same results, it is just that the order in which variables appear will differ.

The mere fact that you want to use difficult suggests to me that there may be problems with your data or models. You may want to look at the gologit2 troubleshooting FAQ at

https://www3.nd.edu/~rwilliam/gologit2/tsfaq.html

If you post code and output, it may be easier to advise you. A replicable example could also be good. See pt #12 of the FAQ.

[Monika Zukowska]

Thank you very much for your quick reply! I have to admit that despite already having looked at this faq file before, only now I realised how I was writing the command completely wrong,
instead

gologit2 y x1 x2 x3, pl(x1) difficult

I was trying to do an equivalent of

gologit2 y pl(x1) x2 x3, difficult

. I fixed this, and now Stata accepts the command (it takes long to compute, nevertheless, so it indeed may not be the solution to my problem).

I am aware that there is a problem with my data and I am in the process of figuring out what to do about it. I am creating similar models for multiple dependent variables. I started by decreasing the number of categories of my DV's (there were to many and some were v. poorly represented), but that solved the problem for some of my DV's. My model also includes interactions and for some of the interaction terms the observation number is relatively small. I am still working on properly identifying the problem, and hopefully finding a solution.