Estimating Probability
Posted: Thu Jun 29, 2017 11:58 am
Hi
I am estimating a model Y=a1x1+a2x2+a3x3+error term
Y is a probability : 0≤Y≤1
And i'm using OLS
The problem is that, any model that I get, after forecasting, I get negative numbers for Y
Some students suggested to use a transformation for Y that is :
Z=log[Y/(1-Y)]
I estimate the model, predict Z, then extract Y.
1/ Is it a correct/good idea ?
2/ Any other suggestion ?
Regards
I am estimating a model Y=a1x1+a2x2+a3x3+error term
Y is a probability : 0≤Y≤1
And i'm using OLS
The problem is that, any model that I get, after forecasting, I get negative numbers for Y
Some students suggested to use a transformation for Y that is :
Z=log[Y/(1-Y)]
I estimate the model, predict Z, then extract Y.
1/ Is it a correct/good idea ?
2/ Any other suggestion ?
Regards