When you choose
menu, you can specify how to eliminate the terms.
- Choose whether to eliminate the least important or most important
- Eliminate unimportant predictors
- Eliminate the least important predictors first to select a
subset of predictors to use for the model. For example, to reduce a set of 500
predictors to the 10 most important predictors. The algorithm removes the least
important predictors sequentially, shows you results that let you compare
models with different numbers of predictors, and produces results for the set
of predictors with the best value of the model selection criterion.
- Eliminate important predictors to assess their impacts
- Eliminate the most important predictors first to assess the
effect on the model. For example, use this option to see the change in the
R2 value as the most important predictors leave the model. The
algorithm removes the most important predictors sequentially, shows you results
that let you evaluate the effect of each important predictor on the accuracy
criterion, and produces results for the model with all the predictors.
- Eliminate K predictors at each step
- Usually, you eliminate 1 predictor at a time. If you have an extremely
large number of predictors and you expect few predictors are very important,
consider a larger value. For example, you can remove more predictors per step
and increase the maximum number of elimination steps to remove more predictors
- Maximum number of elimination steps
- Usually, the maximum number of elimination steps is the number of
reduced models you want to examine, but the algorithm stops early if the model
runs out of predictors. When you increase the number, you usually eliminate a
small number of predictors at each step relative to the number of predictors
and want to continue so that you can see smaller models. For example, you can
remove more predictors per step and increase the maximum number of elimination
steps to remove more predictors faster. Decrease this value to evaluate fewer
- Specify predictors to be removed last
- Specify a subset of predictors to remove after the rest of the
predictors. For example, You have 10 predictors and specify 3 predictors to
remove last. The algorithm removes the other 7 predictors before it considers
any of the 3 predictors that you specify. Usually, you specify predictors to
remove last when you have a special interest in one or more predictors. For
example, you can specify the predictors to remove last so that the algorithm
evaluates a model with only those predictors.
- Display model selection table
- Choose whether to display the results for the training data.
- For test set
- Usually, you display the results for the test set. The algorithm
uses these results to determine which variables to eliminate. The test results
indicate whether the model can adequately predict the response values for new
observations, or properly summarize the relationships between the response and
the predictor variables.
- For test and training sets
- The training results are usually more ideal than the actual
results for new data. The training results are for reference only.