Janssen GB ; Krajnc EI ; Hoekstra JA ; Speijers
GJA
25 p
in Dutch
1991
Toon Nederlands
English Abstract This report describes an alternative statistical method
for the growth of rats in semi-chronic toxicity-experiments. The new method
is compared with the standaard method using the results of the semi-chronic
toxicity experiment of propylgallate. The standard method is analysis of
variance applied to the body-weight after each period, the cumulative
grotwth after each period and the growth within each period followed by
Dunnett's t-test for comparison between dose groups and control. The
alternative method used is curve-fitting. With this method all measurements
of the body-weight in the experimental period for each animal are reduced to
three parameters which describe the curve in time. Disadvantages of the
previously used method are that no account is made for the fact the results
within each period are not independent, that more tables have to be
generated and that a large number of tests have to be done. The large
number of tests together with the dependent results within each period can
lead to illogical patterns by spuriously significant differences. The
curve-fitting method resulted in a curve which adequately described the data
and did not give rise to heterogeneous residuals. The results are
summarized more succinctly in one tabel and of the fitted parameters A and Q
have a good biological interpretation so that direct information is
available over the real growth. Treatment effects on the fitted constants
could be analysed with ANOVA with Dunnett's t-test.
