Home > Statistics Every Writer Should Know > The Stats Board > Discusssion
t-test for comparison of parameters estimates between two logistics models.
Message posted by Vicky on April 17, 2000 at 12:00 AM (ET)
Greeting for all you,
I thank for the aid to them that offers to me.
I have two groups, and in both is the same variable and to be modeled in function of same variable X. The model that I will consider is Logistic:
Y = ko / 1 + b2 e^( -b1 x)
I want to make a t-test to compare if difference between the parameters estimates between groups exists:
H0 : b1 groupA = b1 groupB
and
H0 : ko groupA = ko groupB.
In many books is this t-test, for linear models, but it does not appear much about this test for models nonlinear.
Thank you very much by the information that offers to me about this test.
READERS RESPOND:
(In chronological order. Most recent at the bottom.)
Re: t-test for comparison of parameters estimates between two logistics models.
Message posted by Jack Tomsky on April 19, 2000 at 12:00 AM (ET)
Vicky, you can obtain the (large-sample) covariance matrices of the b1 and k0 estimates by inverting the information matrices. In general, k0 and b1 will be correlated. So, to simultaneously test that the two groups have the same b1 and k0, you end up with an F test. If the test fails, you can then use multiple comparisons to find which parameters are different.
Your $5 contribution helps cover part the $500 annual cost of keeping this site online.