Overparametrised one-way ANOVA — what is and isn't estimable

In a one-way ANOVA model $y_{ij}=\mu+d_i+e_{ij}, j=1,\dots,n_i,\ i=1,\dots,4$ with $E(e_{ij})=0$. For all $j$ and $i$, which of the following statements is true?