Express the following model in matrix form, ie: specify $\beta$ and $X$ so that the model can be written as $Y = X \beta + \epsilon$.

The model $Y_{ij} = \mu_i + \epsilon_i$ where $Y_{ij}$ represents the $j$-th observation observed at level $i$ of a qualitative explanatory variable $T$, for $i=0,...,4$ and $j=1,2$.

I know how to write the vector $\beta$ (just a column of $\mu_0$ to $\mu_4$), but how do you write the matrix X?


Write $\tilde{Y}$ as a vector of length $10,$ by stacking $Y_{01},Y_{02},Y_{11},Y_{12},...,Y_{42}.$ Similarly define $\tilde{\varepsilon}.$ You are now looking for a $10\times 5$ matrix $X$.


$$X=\begin{bmatrix}1&0&0&0&0\\ 1&0&0&0&0\\ 0&1&0&0&0\\ 0&1&0&0&0\\ 0&0&1&0&0\\ 0&0&1&0&0\\ 0&0&0&1&0\\ 0&0&0&1&0\\ 0&0&0&0&1\\ 0&0&0&0&1\end{bmatrix}$$

Then $\tilde{Y}=X\beta+\tilde{\varepsilon}$ is an expression of $Y_{ij}=\mu_i+\varepsilon_{ij}.$

If instead you had $Y_{ij}=\mu_0+\mu_i+\varepsilon_{ij},$ $\tilde{Y}$ and $\tilde{\varepsilon}$ will have length $8.$ Then

$$X=\begin{bmatrix}1&1&0&0\\ 1&1&0&0\\ 1&1&0&0\\ 1&1&0&0\\ 1&0&1&0\\ 1&0&1&0\\ 1&0&0&1\\ 1&0&0&1\end{bmatrix}$$

will be your design matrix so that $\tilde{Y}=X\beta+\tilde{\varepsilon}.$

  • $\begingroup$ I understand that, my problem was with how to construct the design matrix X. I know it is made up of 1's and 0's, and that the first column is just 1's, but how do you determine the other entries? $\endgroup$ – Student_1996 Apr 20 at 21:03
  • $\begingroup$ The first column will not be all $1$'s in the form you wrote. To get $Y_{ij}=\mu_i+\epsilon_{ij},$ take $X_{ij}=1$ and $0$ otherwise. See if this helps. $\endgroup$ – Arnab Auddy Apr 20 at 22:00
  • $\begingroup$ I still don't understand for which cases the ij-th entry is 1? $\endgroup$ – Student_1996 Apr 21 at 8:18
  • $\begingroup$ Sorry I was not clear enough. I have edited the answer. $\endgroup$ – Arnab Auddy Apr 21 at 19:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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