Known fate error message - grouping problem??

[Moynahan]

I'm working on a known fate analysis with 11 intervals and 16 individual covariates. I'd like to code the INP file with 4 groups, corresponding to four study sites. For the simplicity of using the PIM chart to give a baseline definition to the various design matrices where I'll write my models, I would much prefer to use the "known fate group=1" method in the INP file rather than using 3 dummy variables. When I try to run the data, I get an error message: "number of values specified is incorrect for this data type." I believe this is a problem with the grouping text, as I've been able to get a stripped-down version of the INP to run when I remove the grouping text and include the dummy variable coding. I've scoured help files, examples, and the gentle introduction, but can't find specs for precisely how to write the grouping lines in the INP files. Thoughts??

Thanks much,

Brendan Moynahan
Missoula, Montana

[brant]

I'm working on a known fate analysis with 11 intervals and 16 individual covariates. I'd like to code the INP file with 4 groups, corresponding to four study sites. For the simplicity of using the PIM chart to give a baseline definition to the various design matrices where I'll write my models, I would much prefer to use the "known fate group=1" method in the INP file rather than using 3 dummy variables.

Brendan,

take a look at the bottom of page 2-3 and top of page 2-4 in the gentle introduction. i think this should provide some of the info you are after.

basically, if you are using individual covariates, you are going to have to stick w/ dummy variables for the groups. yes, this makes things a little bit messier, but, on the whole, it is not too bad. this also means you will have to get your "hands wet" w/ the matrices.

in general, i have had a hard time in the past with the code that you tried (e.g. "known fate group=1"). Although i dont have a copy of mark on hand at the moment, if you wish to drop the individual covariates and try an inp file like this:

0010101010001011 45 64 23;
0000001010110000 2 4 6 ;
...

where 1st column is cap. history, second in # of indivs. in group 1, 3rd is # of indiv. in group 2, etc. then make sure you set group=3 or whatever when you read in the inp file. this might work (but i would stick w/ dummy variables, particularly for your purpose).

again, no promises because i cannot check on it at the moment.

good luck,
b

[Moynahan]

Thanks to the folks that posted or emailed with replies. The crux of the problem was that the known fate module is not as flexible as the nest survival module (with which I was much more familiar). Where the nest survival module can handle groups (written as "known fate group=x") AND individual covariates in the same INP file, the known fate module can only handle one or the other - not both at the same time.

I got rid of the groups in the INP file and coded them as individual covariates (my four groups: 1000, 0100, 0010, and 0001), then tacked on whatever individual covariates I wanted. This worked fine.