PROBIT model with continous variable?

[sad honk]

Dear STATAforum.
Im reaching out to you because Im struggling to understand my tutor.
He is suggesting that I run a Probit model to overcome the challenge of OLS not adhering to a 0/1 constraint for probability ("LPM problem").
However; my issue is that the dependant variable which I am predicting onto, is a continous variable (a ratio / a share) which holds observations between 0 and 1 (0.2, 0.4, 0.235... so on and so forth).
My understanding is that Probit only predicts onto a dichotomous variable.

Is there any way I feasibly may transform my dependant variable into a dummy? I was thinking about determining the ratio / share as either "in or out", however I have several control variables (independent variables) which affects the dependant variable's value, so its not straight forward for me to just use "margins at" to determine whether an entity is "in our out" of the state which I am trying to find.
Would appreciate any help.

[Nick Cox]

Let's hope that -- given that you are presumably using your real name following https://www.statalist.org/forums/help#realnames, although you didn't get as far as https://www.statalist.org/forums/help#spelling -- your tutor isn't reading this.

Your tutor is right. Although probit was originally designed for (0, 1) outcomes (or dependent variables), it can be applied to continuous proportions.

Degrading continuous proportions to (0, 1) indicator variables -- dummy variables in your terminology -- is not only not needed, it would be throwing away information.

help fracreg

to learn about fracreg probit

[Maarten Buis]

First things first: if you don't understand what your tutor is telling you, then you need to ask your tutor. That is crucial feedback for your tutor that you are withholding from them/her/him. If you want to have a productive working relationship with your tutor you need to ask questions to the tutor and not to an internet forum. You are welcome to ask questions on this forum, but I am telling you that I see it as a symptom of a bigger problem that you probably want to fix. I am usually on the other side of the equation, and the one type of student that annoys me the most is the students who always say that everything is clear, and than it later turns out that that is not the case. My job is to help that student, and by not communicating they make my job impossible. On the other hand, I love working with students who for whatever reason don't get it and ask. It does not have to be a "complete" or "right" question; a big part of the consulting process is to find the "real" question. So first advise: talk to your tutor.

Second, the statalist has a long tradition that is a bit different from other forums, in that we use our real names. We have found that this encouraged polite and professional communication on this forum, so we tend to invite new members to contributed to that tradition, and also use their real name. You can request the administrator to change your name by clicking the "contact us" button at the bottom of this page.

Finally, we get to your question. The answer is that you do not want to turn your dependent variable into a dummy (or binary or indicator) variable. Instead, you want to use models that are intended for fractional outcomes, and there is one with a "probit flavor" to it. So I assume that that is what your advisor meant. See help fracreg. Don't forget to also look at the pdf-manual entry for that command: the pdf manual tends to contain more information than the help-file. A link to the pdf-manual can be found at the top of the help-file. An overview of this topic can also be found here: http://www.maartenbuis.nl/publications/prop.html