This book is in Open Review. We want your feedback to make the book better for you and other students. You may annotate some text by selecting it with the cursor and then click the on the pop-up menu. You can also see the annotations of others: click the in the upper right hand corner of the page

7.1 Hypothesis Tests and Confidence Intervals for a Single Coefficient

We first discuss how to compute standard errors, how to test hypotheses and how to construct confidence intervals for a single regression coefficient \(\beta_j\) in a multiple regression model. The basic idea is summarized in Key Concept 7.1.

Key Concept 7.1

Testing the Hypothesis \(\beta_j = \beta_{j,0}\)
Against the Alternative \(\beta_j \neq \beta_{j,0}\)

  1. Compute the standard error of \(\hat{\beta_j}\)
  2. Compute the \(t\)-statistic, \[t^{act} = \frac{\hat{\beta}_j - \beta_{j,0}} {SE(\hat{\beta_j})}\]
  3. Compute the \(p\)-value, \[p\text{-value} = 2 \Phi(-|t^{act}|)\]
where \(t^{act}\) is the value of the \(t\)-statistic actually computed. Reject the hypothesis at the \(5\%\) significance level if the \(p\)-value is less than \(0.05\) or, equivalently, if \(|t^{act}| > 1.96\). The standard error and (typically) the \(t\)-statistic and the corresponding \(p\)-value for testing \(\beta_j = 0\) are computed automatically by suitable R functions, e.g., by summary.

Testing a single hypothesis about the significance of a coefficient in the multiple regression model proceeds as in in the simple regression model.

You can easily see this by inspecting the coefficient summary of the regression model

\[ TestScore = \beta_0 + \beta_1 \times size \beta_2 \times english + u \]

already discussed in Chapter 6. Let us review this:

model <- lm(score ~ size + english, data = CASchools)
coeftest(model, vcov. = vcovHC, type = "HC1")
#> t test of coefficients:
#>               Estimate Std. Error  t value Pr(>|t|)    
#> (Intercept) 686.032245   8.728225  78.5993  < 2e-16 ***
#> size         -1.101296   0.432847  -2.5443  0.01131 *  
#> english      -0.649777   0.031032 -20.9391  < 2e-16 ***
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

You may check that these quantities are computed as in the simple regression model by computing the \(t\)-statistics or \(p\)-values by hand using the output above and R as a calculator.

For example, using the definition of the \(p\)-value for a two-sided test as given in Key Concept 7.1, we can confirm the \(p\)-value for a test of the hypothesis that the coefficient \(\beta_1\), the coefficient on size, to be approximately zero.

# compute two-sided p-value
2 * (1 - pt(abs(coeftest(model, vcov. = vcovHC, type = "HC1")[2, 3]),
            df = model$df.residual))
#> [1] 0.01130921

Key Concept 7.2

Confidence Intervals for a Single Coefficient in Multiple Regression

A \(95\%\) two-sided confidence interval for the coefficient \(\beta_j\) is an interval that contains the true value of \(\beta_j\) with a \(95 \%\) probability; that is, it contains the true value of \(\beta_j\) in \(95 \%\) of all repeated samples. Equivalently, it is the set of values of \(\beta_j\) that cannot be rejected by a \(5 \%\) two-sided hypothesis test. When the sample size is large, the \(95 \%\) confidence interval for \(\beta_j\) is \[\left[\hat{\beta_j}- 1.96 \times SE(\hat{\beta}_j), \hat{\beta_j} + 1.96 \times SE(\hat{\beta_j})\right].\]