Ordinal logistic regression data upload and ologitgof command

[vilma manjani]

Hi I am using Stata to run an ordinal logistic regression about ordinal data. my dependent variable is ordinal with 1 indicating not important and 5 important. The participants rated 1 to 5 instead of not important to important. the same for independent variables. so the data set has numbers from1 to 5 in its substance. on the other hand i coded the controls myself before uploading to stata,some are dummy some ordinal. so i was just asking myself if it is fine leaving it like this, will ologit still provide the right results, or i should have them uploaded all in categorical form and stata does the allocation itself?

in addition, another question: i have been trying to run the ologitgof to test for the goodness of fit for the data but when i specify in thevarlist all the variables, stata returns that doesn't have memory for that. when i remove some variables the p value in the chi2 part returns a dot. What does it mean and how do i fix this. is it okay if i report only the HL test and lipsitz test without reporting PR?

Lastly, is it fine to use pwcorr to test correlation between these variables?

[Salvatore Viola]

Dear Vilma,

Thank you for your question. The data should be interpreted fine by the ologit command as far as I know. Still, your question remains a little unclear since you have not shared any of your data with the forum. Without having access to a portion of your data through the dataex command (see help dataex in Stata) it is quite difficult for members of the forum to identify and hopefully resolve your problem.

Best,
Salvatore

[Richard Williams]

Welcome to Statalist Vilma.

I'm not sure I understand your questions. ologit, like most Stata commands, is going to treat independent variables as continuous unless you tell it otherwise via factor variable notable, e.g. if relig is coded something like 1 = Catholic, 2 = Protestant, 3 = Jewish, 4 = other, your command should probably be something like

ologit y x i.relig

Perhaps you are wondering whether it is legitimate for ordinal independent variables to be treated as continuous. For a discussion, see

https://www3.nd.edu/~rwilliam/xsoc73...ndependent.pdf

or, better yet, if your library gives you access, see

https://methods.sagepub.com/Foundati...dent-variables

As Salvatore suggests, providing us with an extract from your data, and also showing us your exact commands and Stata's response, may help us better advise you.