[Nahed Eddai]

Dear Richard,
Thank you for your answer. My dependent variable takes the amount of tokens the player can give to a public account in the frame of a public goods game experiment. So each subject is endowed with 200 tokens and has to choose one of the 5 possible contributions to give to the public account.
I hope that it is more clear now.
Best regards

[Nick Cox]

I see two possibilities here: ordered logit or probit on the one hand, and Poisson regression on the other. OLS is an estimation method and not a model, but the implied alternative of plain linear regression sounds to me the least suitable choice, if only because (and not only because) the possibility of negative predictions is quite stark.

[Richard Williams]

This is an unusual problem. I agree with Nick that ologit or oprobit sounds best.