# Forecasting Using the Linear Regression Model

5 Pages Posted: 23 Jun 2009

See all articles by Phillip E. Pfeifer

## Phillip E. Pfeifer

University of Virginia - Darden School of Business

### Abstract

This technical note explains how to use a linear regression model to construct a point-and-interval forecast of an uncertain dependent variable.

Excerpt

UVA-QA-0327

FORECASTING USING THE LINEAR REGRESSION MODEL

A previous note, “Linear Model Building” (UVA-QA-0293), presented a procedure for building a model that will provide reliable forecasts of the relevant uncertain quantity (the Y-variable), given knowledge of one or more influential factors (X-variables). For a model to be useful in this forecasting role, it must (1) make sense (i.e., be consistent with the model-builder's beliefs about the way things work), (2) be simple (i.e., not contain extraneous, insignificant terms), and (3) meet the assumptions underlying the model. Given that such a model has been built, exactly how do you use the model to forecast?

Point Forecast

The first step in using the chosen model to forecast is an easy one. Simply substitute the relevant X-value(s) into the fitted equation

Y-hat = a-hat + b-hat(X)

to calculate Y-hat, the point forecast of Y for the given value of X.

As an example, recall that in a previous note, “Introduction to Least-Squares Modeling” (UVA-QA-0500), we used 20 pairs of observations of Sunnyvale's bid per batch of transistors, the Y-variable, and the corresponding estimated number of good transistors in each batch, the X-variable, to fit the model

. . .

Keywords: forecasting, regression analysis, statistics, uncertainty

Suggested Citation

Pfeifer, Phillip E., Forecasting Using the Linear Regression Model. Darden Case No. UVA-QA-0327, Available at SSRN: https://ssrn.com/abstract=1422916

## Paper statistics 