Chapter Contents |
Previous |
Next |

Logistic Regression |

Choose Tables:Type III (LR) Tests. |

**Figure 16.6:** Tables Menu

This displays the table shown in Figure 16.7.

The *p*-values for **TEMP** and **CELL** are relatively large,
suggesting these effects could be removed.
Although the numbers are different, the same conclusions
would be reached from the corresponding **Wald** tests.
In the Fit Variables dialog, follow these steps
to request a new model with **TEMP** removed.

Select TEMP in the effects list, then click the Remove button. |

**TEMP** disappears from the effects list.

Click on Apply, and a new fit window appears,
as shown in Figure 16.8. |

Choose Tables:Type III (LR) Tests in the new fit window. |

This displays a **Type III (LR) Tests** table in the window.

The *p*-value for **CELL** in the LR test suggests that this effect
could also be removed.

Click on the variable CELL in the effects list in the Fit Dialog. |

Then click on **Remove**. **CELL** disappears from the effects list.

Click on Apply, and a new Fit
window appears, as shown in Figure 16.10. |

Since the new model contains only one **X** variable, the fit window
displays a plot of **REMISS** versus **CELL**.

Using the **Apply** button, you have quickly
created three logistic regression models.
Logistic regression is only one special
case of the generalized linear model.
Another case, Poisson regression,
is described in the next chapter.

Related Reading | Generalized Linear Models, Chapter 39. |

Chapter Contents |
Previous |
Next |
Top |

Copyright © 1999 by SAS Institute Inc., Cary, NC, USA. All rights reserved.