Hello,

I estimated the logit model, but I need to plot the distribution (PDF) underlying the logit model. Could you please provide some help on this?

Thanks!

## 1 Answer

0

Hello,

If I am interpreting your question correctly you want to be able to graph the underlying logistic function y = e^(b0+xb)/(1+e^(b0 + xb)). I am not sure how you have estimated your logit model but I can provide a general explanation. Suppose we have estimated the logit model of the dependent variable, *y*, on the dependent variable, *x*. The model yields two estimates, an intercept, *b0* and a coefficient, *b*. In order to create your graph, you need to first generate y_pred = b0 + xb. However, for the sake of the graph, the y_pred should be created using a sorted x.

x_sorted = sortc(x,1);

y_pred = b0 + x_sorted.*b;

Next, *y_pred* needs to be transformed logistically:

y_gr = exp(y_pred)./(1+ exp(y_pred));

This can now be directly graphed using the **plotXY** function.

plotXY(x_sorted, y_gr);

If you provide more information regarding how you estimated the logit model or provide code, I would be happy to provide a more detailed and specific explanation.

## Your Answer

## 1 Answer

Hello,

If I am interpreting your question correctly you want to be able to graph the underlying logistic function y = e^(b0+xb)/(1+e^(b0 + xb)). I am not sure how you have estimated your logit model but I can provide a general explanation. Suppose we have estimated the logit model of the dependent variable, *y*, on the dependent variable, *x*. The model yields two estimates, an intercept, *b0* and a coefficient, *b*. In order to create your graph, you need to first generate y_pred = b0 + xb. However, for the sake of the graph, the y_pred should be created using a sorted x.

x_sorted = sortc(x,1);

y_pred = b0 + x_sorted.*b;

Next, *y_pred* needs to be transformed logistically:

y_gr = exp(y_pred)./(1+ exp(y_pred));

This can now be directly graphed using the **plotXY** function.

plotXY(x_sorted, y_gr);

If you provide more information regarding how you estimated the logit model or provide code, I would be happy to provide a more detailed and specific explanation.