The parameter T in the model function
P = 1/[r k (T - t)]1/r
is obviously important. It is just as obviously not directly observable from measurements or estimates of population. However, we can take logs of both sides of this equation to find the equivalent form
log P = (-1/r) [log (rk) + log (T - t)]
If this model fits the data (and we have seen some evidence that it does), then we should find that log P is a linear function of log (T - t).