About rank order logit model

[Pratikshya Silwal]

Hi there,
I am having trouble running the rank order logit model for my data. I used fish ranking as a dependent variable and various other independent variables. I used the seven different fish species as a group variable and those seven different fish species dummy variables were also used as independent variables for determining preference for various fish. Their rank is in the form of a Likert scale 1 being least preferred to 5 being highly preferred. While running the analysis I got results omitting all the coefficients mentioning no within-species variance within fish species. I have posted the code and the result I got. Could anyone suggest why all those values are omitted and are there any way to run this model? Thank you

[Hong Il Yoo]

Please note that -group()- variable must identify distinct choice occasions rather than distinct alternatives. Let n denote individuals, j denote alternatives, and t denote choice occasions. Your -group()- variable must take different values for different {n, t} pairs. Based on your description, it sounds like your -group()- variable takes different values for different {j} instead. I would like to suggest that you -browse- the example data file for -help rologit- and look at what its -group()- variable looks like. Once your -group()- variable has been selected correctly, you can -browse- your own data and see if the omitted regressors show within-group variations. If they do not show within-group variations, you are attempting to estimate an unidentified model: Only differences in utility matter to observed choice (and ranking) behaviour.