What is wrong with my Tobit regression?

[d6k5d3]

I have dependent variable Y which has both negative and positive values. I want to test the impact of some independent variables (Xs) on the magnitude of Y. Hence, I chose the Tobit model. I took the absolute value of Y (abs_Y), which is my new dependent variable for my Tobit regression. I wrote the following code:

proc qlim data= YYYZZZ;
model abs_Y= X1-X34; nloptions maxiter= 500;
endogenous abs_Y ~ censored (lb= 0);
run;

The output is what dumbfounded me.

I need your advice as to what may have gone wrong and how to get rid of this. SAS also generated graphs for predicted abs_Y by regressors. I am also not sure how to interpret the graphs given I have the above estimates. Let me share the graphs here.

Much thanks for all your insights.

[Reeza]

Your output doesn't match your code. Why is the variable in the output named ABS_JUMP but your code shows ABS_Y?

---

@d6k5d3 wrote:

I have dependent variable Y which has both negative and positive values. I want to test the impact of some independent variables (Xs) on the magnitude of Y. Hence, I chose the Tobit model. I took the absolute value of Y (abs_Y), which is my new dependent variable for my Tobit regression. I wrote the following code:

proc qlim data= YYYZZZ;
model abs_Y= X1-X34; nloptions maxiter= 500;
endogenous abs_Y ~ censored (lb= 0);
run;

The output is what dumbfounded me.

I need your advice as to what may have gone wrong and how to get rid of this. SAS also generated graphs for predicted abs_Y by regressors. I am also not sure how to interpret the graphs given I have the above estimates. Let me share the graphs here.

Much thanks for all your insights.

---

[d6k5d3]

Oh! That's for just quick posting! The coding is same!

[Reeza]

I can only comment on what you show. Please show your actual code.

[d6k5d3 · Accepted solution]

Abs_Jump is nothing but abs_Y. I can make sure that representation of the variables in the coding is accurate.