EXCEL: Multiple Regression

This page gives information on how to

There is little extra to know beyond regression with one explanatory variable.
The main addition is the F-test for overall fit.

Here Tools | Data Analysis | Regression is used.
One can also use Excel functions LINEEST and TREND, see Excel: Two Variable Regression using Excel Functions
 

DATA

Consider the following data on the number of cars in a household (first column) and the number of persons in the household (second column).
 

   1     1 
   2     2 
   2    3
   2    4
   3    5

Type the data into cells A1:A5 and cells B1:B5.
Then give headers in a new first row of CARS and HH SIZE to the two columns.
Then create a new variable in cells C2:C6, cubed household size as a regressor, and give in C1 the heading CUBED HH SIZE. (It turns out that squared hh size has a coefficient of exactly 0.0 so I use the cube).

The spreadsheet cells A1:C6 should look like:
 

CARS HH SIZE CUBED HH SIZE
   1    1      1
   2    2      8
   2    3     27
   2    4     64
   3    5    125

RUN THE REGRESSION

The population regression model is

      y = b1 + b2*x + b3*z + u

where here z = x^3 and the error term u has mean 0 and variance sigma-squared.

We wish to estimate the regression line

     y = b1 + b2*x + b3*z

Do this by Tools / Data Analysis / Regression.
The only change over one-variable regression is to include more than one column in the Input X Range.

A relatively simple form of the command (with labels) is
 

Input Y Range a1:a6
Input X Range b1:c6
Labels Tick
Output range a8

and then hit OK.

Complete output for this example is given at the bottom.
Here we go through the output in detail, considering both data summary and statistical inference.
 

REGRESSION STATISTICS

This is the following output. Of greatest interest is R Square.
 

    Explanation
Multiple R 0.895828 R = square root of R^2
R Square 0.802508 R^2 = coefficient of determination
Adjusted R Square 0.605016 Adjusted R^2 used if more than one x variable
Standard Error 0.444401 This is the sample estimate of the st. dev. of the error u
Observations 5 Number of observations used in the regression (n)

The above gives the overall goodness-of-fit measures:
      R-squared = 0.8025
      Correlation between y and y-hat is 0.8958 (when squared gives 0.8).
      Adjusted R-squared = R^2 - (1-R^2)*(k-1)/(n-k) = .8025 - .1975*2/2

The standard error here refers to the estimated standard deviation of the error term u.
It is sometimes called the standard error of the regression. It equals sqrt(SSE/(n-k)).
It is not to be confused with the standard error of y itself (from descriptive statistics) or with the standard errors of the regression coefficients given below.
 

ANOVA

An ANOVA table is given. This is often skipped.
 

  df SS MS F Significance F
Regression 2 1.6050 0.8025 4.0635 0.1975
Residual 2 0.3950 0.1975    
Total 4 2.0      

The above ANOVA (analysis of variance) table splits the sum of squares into its components.

Total sums of squares
= Residual (or error) sum of squares + Regression (or explained) sum of squares.

Thus Sum (y_i - ybar)^2 = Sum (y_i - yhat_i)^2 + Sum (yhat_i - ybar)^2.

The column labeled F gives the overall F-test of H0: b2 = 0 and b3 = 0 versus Ha: at least one does not equal zero.
Aside: Excel computes F this as:
F = [Regression SS/(k-1)] / [Residual SS/(n-k)] = [1.6050/2] / [.39498/2] = 4.0635.

The column labeled significance F has the associated P-value. Since it is greater than 0.05 at the 5% siginificance level we do not reject H0.
 

KEY REGRESSION OUTPUT

The regression output of most interest is the following table of coefficients and associated output:
 

  Coefficient St. error t Stat P-value Lower 95% Upper 95%
Intercept 0.89655 0.76440 1.1729 0.3616 -2.3924 4.1855
HH SIZE 0.33647 0.42270 0.7960 0.5095 -1.4823 2.1552
CUBED HH SIZE 0.00209 0.01311 0.1594 0.8880 -0.0543 0.0585

Let bj denote the population coefficient of the jth regressor (intercept, HH SIZE and CUBED HH SIZE).

Then

A simple summary of the above output is that the fitted line is

     y = 0.8966 + 0.3365*x + 0.0021*z
 

TEST OF STATISTICAL SIGNIFICANCE

The coefficient of HH SIZE has estimated standard error of 0.4227, t-statistic of 0.7960 and p-value of 0.5095.
It is therefore statistically insignificant at 5%.

The coefficient of CUBED HH SIZE has estimated standard error of 0.0131, t-statistic of 0.1594 and p-value of 0.8880.
It is therefore statistically insignificant at 5%.

There are 5 observations and 3 regressors (intercept and x) so we use t(5-3)=t(2).
 

GOODNESS OF FIT

A measure of the fit of the model is R^2=0.803, which means that 80.3% of the variation of y_i around ybar is explained by the regressor x_i.
The standard error of the regression is 0.365148.
 

TEST HYPOTHESIS ON A REGRESSION PARAMETER

Here we consider test of zero slope parameter.
This will reproduce the results for statistical significance. We could change to testing other values.

Example: Test H0: b2 = 0 against Ha: b2 not equal to 0, at significance level alpha = .05.

Do not reject H0 as p-value = 0.5095 > 0.05.
Equivalently, so not reject H0 as t2 = 0.7960 < t_.025(2) = 4.303 (from tables where here n-k = 4-3 = 2).

Aside: Excel calculates t2 = b2 / se2 = 0.3365/0.4227
           and p-value as Pr(|t| > 0.5095 | t is t(2))
 

CONFIDENCE INTERVAL FOR REGRESSION PARAMETER

95% confidence interval for coefficient of HH SIZE is from Excel output (-1.4823, 2.1552).
 

OVERALL TEST OF SIGNIFICANCE OF THE REGRESSION PARAMETERS

We test H0: b2 = 0 and b3 = 0  against  Ha: at least one of b2 and b3 is not equal to 0.
From the ANOVA table the F-test statistic is 4.0635 with p-value of 0.1975.
Since the p-value is not less than 0.05 we do not reject the null hypothesis that the regression parameters are zero at significance level 5%.
Conclude that the parameters are jointly statistically insignificant at significance level 5%.
 

PREDICTED VALUE OF Y GIVEN REGRESSORS

Consider case where x = 4 in which case z = x^3 = 4^3 = 64.

yhat  = b1 + b2*x + b3*z  =  0.88966 + 0.3365*4 + 0.0021*64 = 2.37006.

This is also the predicted value of the conditional mean of y given x=4 and z=64.
 

OTHER REGRESSION OUTPUT

Other useful options of the regression command include:
Confidence Level other than 95% gives confidence intervals for b1 and b2 at level other than 95%.
Line Fit Plots plots the fitted and actual values yhat_i and y_i against x_i.
These are similar to what is discussed in bivariate regression.
 

LIMITATIONS

Excel will only permit up to 16 regressors.

Excel requires that all the regressor variables be in adjoining columns.
You may need to move columns to ensure this.
e.g. If the regressors are in columns b and d you need to copy at least one of columns b and d so that they are adjacent to each other.
 

EXCEL OUTPUT FOR THIS EXERCISE