Consider the following model $y_i = \beta_0+ \beta_1x_{i1}+ \beta_2x_{i2}+ \epsilon_i$ where $\epsilon_i \overset{iid}{\sim} N(0, \sigma^2)$.

How do I test the null hypothesis that $\beta_1 + 10\beta_2 = 0$?


1 Answer 1


Quoting from Wikipedia:

The variance of a sum of two random variables is given by $$\operatorname{Var}(aX+bY)=a^2\operatorname{Var}(X)+b^2\operatorname{Var}(Y)+2ab\operatorname{Cov}(X,Y)$$

where $\operatorname{Cov(⋅, ⋅)}$ is the covariance.

In your case, substitute $a = 1$, $b=10$, $X = \beta_1$, and $Y=\beta_2$. Statistical software will typically provide a variance-covariance matrix for the coefficients of a linear regression model (although you might have to look into the documentation to figure out how to find it). The individual coefficient variances are along the diagonal, the $(\beta_1,\beta_1)$ and $(\beta_2,\beta_2)$ entries respectively. The $(\beta_1,\beta_2)$ entry of that matrix provides the desired covariance value.

The square root of that variance of the weighted sum of the coefficients is the estimate of its standard deviation. With a sufficiently large number of cases, if $\beta_1+10\beta_2$ is more than 1.96 standard deviations away from 0, then you have a significant difference at p < 0.05 in a 2-sided z-test (equivalent to a Wald test). With a smaller number of cases you could use a t-test .


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.