Hi Masud,

Indeed, it might be more interesting to use a logistic regression in this context. What you can do to check this latter point is to compute an OLS regression, compute the MSE, and compare it to the MSE that one can get using a logistic regression.

Anyway, logistic regression family is probably is good option, the remaining question is "which type should I use".

Binary dependent variable : No

Ordered : It depends on your dependent variable, e.g. Is there an idea of an "increase in something" ? Or is your variable somehow comparable to a variable "what is in your pocket : 1/ An apple ; 2/ A peach 3/ A strawberry , etc... , which is non ordinal variable.

Ordered logistic regression works on ordinal variables, not on non-ordinal variables.

Non-ordered : Then, consider using multinomial logistic regression. https://en.wikipedia.org/wiki/Multinomial_logistic_regression

Considering that this issue has already been adressed recently in an other RG question, please find below a link for more details :

https://www.researchgate.net/post/How_to_use_logistic_regression_analysis_for_more_than_two_class_problem#view=57726a923d7f4b0519484a0b

I hope this helps,

Best regards,

Hi

Ordered probit and logit models can handle your problem.. It is not necessary to have ordered numbers of positive values as stated in your question, they can be negative values too, but only the  increment in the numbers has to be one unit.

Dear Geoffrey and Prof. Ibrahim,

Think you so much for your insightful answers. Your answers are very helpful.

Sincerely,

Masud