In one-way ANOVA, $c\mu+\sum d_i\alpha_i$ is estimable iff $\sum d_i=c$

Under completely randomized design with model $E\,y_{ij}=\mu+\alpha_i+\epsilon_{ij}$, $i=1,\dots,p$, $j=1,\dots,n_i$, $\epsilon_{ij}\sim N(0,\sigma^2)$ and independently distributed, a parametric function $c\mu+\sum_{i=1}^pd_i\alpha_i$ where $c$ and $d_1,\dots,d_p$ are known constants is estimable if and only if: