Knowing what to make of data can be very difficult. A linear model allows an analyst to predict the value of an important but unknown quantity with by plugging known values into a linear equation. Different combinations of variables have different levels of effectiveness in predicting the values of the response variable. One way to measure the effectiveness of a model is SSE, or sum of squared error. Deciding which predictors to include in a model and which to leave out involves a trade-off between how much that predictor decreases the SSE (its added sum of squares) on the one hand and the risk of over-fitting the model by adding too many predictors on the other. This program implements an automatic model selector that employs the forward selection algorithm. This provides a great starting point for developing a model and eliminates the tedium of searching manually.
No comments:
Post a Comment