# Problems with predicted probability logit/probit

#### LiukMarco

##### New Member
Hi

I have to process a probit regression with 10 variables, 2 continuous, 6 dummies and 2 categorical.
So I process xi: probit y x1 x2 x3 x4 x5 x6 x7 x8 i.x9 i.x10

After I want to have the predected probability for the variables' coefficients. Which command shall I use?
Margins (but it doesn't work with the first 2 continuous variables)?

Thanks

#### bukharin

##### RoboStataRaptor
1. If you're using Stata 11 or 12 you don't need, and shouldn't be using, the -xi:- prefix
2. What do you mean by the "predicted probability for the variables' coefficients" - what exactly are you trying to do?

#### LiukMarco

##### New Member
In order,
1.I have stata 11; I use xi: only because I use it to produce the dummies (the two categoricl have 15 and 35 cateogories each); I could use tabulate, it's the same. Why do you say I shoudl not use xi. ?
2. Sorry for that, it was kind of confused. Since the coefficient calculated through the probit give me only the "direction" of the change in the probability that Y=1, I would like also to know the real change that a variation of each independent variable produces on the probability (what a simple regression model shows me since the beginning, to be clear)

#### bukharin

##### RoboStataRaptor
1. You don't need to use -xi:- to create the dummies in Stata 11 or 12. Try it and see.
2. You can use -margins-... but it will get confused by -xi:- so fix that first

I think you're basically after a marginal effect, and you need to use something like:
margins, dydx(x1)

#### LiukMarco

##### New Member
1. So you're telling me that using xi: probit y x1 x2 x3 x4 x5 x6 x7 x8 i.x9 i.x10 is useless. So, how can I produce all the dummy variables for the last two? Through "tabulate"? Since I have a dataset LONG, I thought xi could be more simple.
2. You have underlined the problem with margins. But there is also another one: if I use the normal probit regression (without xi), I put the i.x for the dummy variables. In this case Stata gives me an error about the first two, continuous, telling me that are not included (if I remember correctly the are not included in b(e) ).
I tried to use mfx compute, dxdy at(mean), but I don't know how much I am right about it, I have never used it before.