Categorical dependent variable?

[Marry Lee]

Dear all,


My question is model related. I examine the effect of a policy on the performance of two types of firms : type A and type B

• I run a binary dependent variable model and I find that the policy increased the performance of type A.

Now, type A firms have 2 sub-types, I want to examine if the policy increased the performance of sub-type 1 compared to sub-type 2. I run a binary dependent variable model where Y = 1 if sub-type 1 and Y= 0 if sub-type 2.

I am afraid that because the policy affected type A firms hence both sub-types 1 and 2, results of this second model are not reliable. Can you please tell me what you think of this problem? How do I do to test the effect of the policy on sub-type 1 while accoung for an effect on sub-type 2.

All the best.

[George Ford]

Your model looks backwards.

I would think the model might be:

Performance = b0 + b1*TypeA

or,

Performance = a0 + a1*TypeAsub1 + a2*TypeAsub2

Then you can ask whether a1==a2 for TypeA firms.

[Marry Lee]

George Ford Thanks for your answer.

[George Ford]

Or,

Performance = a0 + a1*TypeA + a2*TypeAsub2

a2 will tell you if there's a difference.