The statistical distribution of size at age is usually modeled as normal; however, no single error structure is used universally when fitting the VBGF. Multiplicative error is assumed most frequently, given the observation that variability of size at age typically increases with age. Additive error is sometimes preferred in cases where variability seems relatively constant with age, for example, when modeling narrow age ranges.
Several modifications to the VBGF allow seasonal variation in growth. One example is due to Hoenig and Hanumara:
The scale of seasonal variation is determined by 6; the starting point of the seasonal cycle, by a1. Growth becomes negative seasonally if 6 >1.
In fitting the VBGF, estimates of k and Lm tend to be highly correlated, which makes comparison of growth rates difficult between populations or over time. Hotelling's T2 statistic has been suggested for testing such comparisons.
The Gompertz model is often considered superior to the VBGF in modeling growth of young fish. Its parame-trization is similar to that of VBGF:
where Wm is the asymptotic weight and k the growth coefficient. The same form is used in modeling weight or length.
Was this article helpful?