*45*

In statistics, we fit regression models for two reasons:

**(1)** To *explain *the relationship between one or more explanatory variables and a response variable.

**(2) **To *predict *values of a response variable based on the values of one or more explanatory variables.

When our goal is to **(2) ***predict* the values of a response variable, we want to make sure that we’re using the best possible regression model to do so.

One metric that we can use to find the regression model that will make the best predictions on new data is the **PRESS Statistic**, which stands for the “**P**redicted **RE**sidual **S**um of **S**quares.”

It is calculated as:

**PRESS = Σ(e _{i} / (1-h_{ii}))^{2}**

where:

**e**The i_{i}:^{th}residual.**h**A measure of the influence (also called “leverage”) of the i_{ii}:^{th}observation on the model fit.

Given several regression models, the one with the lowest PRESS should be selected as the one that will perform best on a new dataset.

The following example shows how to calculate the PRESS statistic for three different linear regression models in R.

**Example: Calculating the PRESS Statistic**

Suppose we have a dataset with three explanatory variables, x_{1}, x_{2}, and x_{3}, and one response variable y:

data data.frame(x1 = c(2, 3, 3, 4, 4, 6, 8, 9, 9, 9), x2 = c(2, 2, 3, 3, 2, 3, 5, 6, 6, 7), x3 = c(12, 14, 14, 13, 8, 8, 9, 14, 11, 7), y = c(23, 24, 15, 9, 14, 17, 22, 26, 34, 35))

The following code shows how to fit three different regression models to this dataset using the **lm()** function:

model1

The following code shows how to calculate the PRESS statistic for each model.

#create custom function to calculate the PRESS statistic PRESS function(model) { i residuals(model)/(1 - lm.influence(model)$hat) sum(i^2) } #calculate PRESS for model 1 PRESS(model1) [1] 590.2197 #calculate PRESS for model 2 PRESS(model2) [1] 519.6435 #calculate PRESS for model 3 PRESS(model3) [1] 537.7503

It turns out that the model with the lowest PRESS statistic is model 2 with a PRESS statistic of **519.6435**. Thus, we would choose this model as the one that is best suited to make predictions on a new dataset.

**Additional Resources**

Introduction to Simple Linear Regression

What is a Parsimonious Model?

What is a Good R-squared Value?