14.03.2014 Views

Basic Analysis and Graphing - SAS

Basic Analysis and Graphing - SAS

Basic Analysis and Graphing - SAS

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

222 Performing Simple Logistic Regression Chapter 7<br />

The Logistic Report<br />

Table 7.1 Description of the Whole Model Test (Continued)<br />

–LogLikelihood<br />

Chi-Square<br />

Prob>ChiSq<br />

Rsquare (U)<br />

AICc<br />

BIC<br />

Observations<br />

(sometimes called<br />

Sum Wgts)<br />

Measures variation, sometimes called uncertainty, in the sample.<br />

Full (the full model) is the negative log-likelihood (or uncertainty) calculated<br />

after fitting the model. The fitting process involves predicting response rates<br />

with a linear model <strong>and</strong> a logistic response function. This value is minimized<br />

by the fitting process.<br />

Reduced (the reduced model) is the negative log-likelihood (or uncertainty)<br />

for the case when the probabilities are estimated by fixed background rates.<br />

This is the background uncertainty when the model has no effects.<br />

The difference of these two negative log-likelihoods is the reduction due to<br />

fitting the model. Two times this value is the likelihood-ratio Chi-square test<br />

statistic.<br />

The likelihood-ratio Chi-square test of the hypothesis that the model fits no<br />

better than fixed response rates across the whole sample. It is twice the<br />

–LogLikelihood for the Difference Model. It is two times the difference of<br />

two negative log-likelihoods, one with whole-population response<br />

probabilities <strong>and</strong> one with each-population response rates.<br />

For more information, see “Statistical Details for the Whole Model Test<br />

Report” on page 233.<br />

The observed significance probability, often called the p value, for the<br />

Chi-square test. It is the probability of getting, by chance alone, a Chi-square<br />

value greater than the one computed. Models are often judged significant if<br />

this probability is below 0.05.<br />

The proportion of the total uncertainty that is attributed to the model fit.<br />

To test that the factor variable has no effect on the response, look at the<br />

difference between the following:<br />

• the log-likelihood from the fitted model<br />

• the log-likelihood from the model that uses horizontal lines<br />

For more information, see “Statistical Details for the Whole Model Test<br />

Report” on page 233.<br />

The corrected Akaike Information Criterion.<br />

The Bayesian Information Criterion.<br />

The total sample size used in computations. If you specified a Weight<br />

variable, this is the sum of the weights.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!