Whenry
Nov23-11, 02:48 PM
Hi all,
I would like some help understand the covariance and correlation matrix as it pertains to estimated parameters. I am performing logistic regression with just one independent variable and a constant term.
I am wondering if someone can help me understand the information these matrices are telling me. I found a lot of information about covariance and correlation between two variables, but not a lot on parameters. If someone could also help me understand how these matrices are calculated, that would be great.
This is what I have so far, please correct me where I am incorrect
the variance-covariance matrix describes how each beta could variance, either on it's own, or together with another parameter and have the model still produce the same (or nearly the same?) values. So for the values below, b0 = -2.4030 could vary +/-0.0036 and we would get estimated y's that are statistically the same, where Y = b0 + b1X.
the correlation matrix describes how you can change two variables together and produce the same fit? in this example, you could increase b0 by 1 and reduce b1 by -0.9035
betas
-2.4030
0.0201
covariance matrix
0.0035526 -3.4778e-05
-3.4778e-05 4.1704e-07
correlation matrix
1.0000 -0.9035
-0.9035 1.0000
I would like some help understand the covariance and correlation matrix as it pertains to estimated parameters. I am performing logistic regression with just one independent variable and a constant term.
I am wondering if someone can help me understand the information these matrices are telling me. I found a lot of information about covariance and correlation between two variables, but not a lot on parameters. If someone could also help me understand how these matrices are calculated, that would be great.
This is what I have so far, please correct me where I am incorrect
the variance-covariance matrix describes how each beta could variance, either on it's own, or together with another parameter and have the model still produce the same (or nearly the same?) values. So for the values below, b0 = -2.4030 could vary +/-0.0036 and we would get estimated y's that are statistically the same, where Y = b0 + b1X.
the correlation matrix describes how you can change two variables together and produce the same fit? in this example, you could increase b0 by 1 and reduce b1 by -0.9035
betas
-2.4030
0.0201
covariance matrix
0.0035526 -3.4778e-05
-3.4778e-05 4.1704e-07
correlation matrix
1.0000 -0.9035
-0.9035 1.0000