Neill's lack-of-fit test for dose-response models
'neill.test' provides a lack-of-fit test for non-linear regression models. It is applicable both in cases where there are replicates (in which case it reduces to the standard lack-of-fit test against an ANOVA model) and in cases where there are no replicates, though then a grouping has to be provided.
neill.test(object, grouping, method = c("c-finest", "finest", "percentiles"), breakp = NULL, display = TRUE)
object |
object of class 'drc' or 'nls'. |
grouping |
character or numeric vector that provides the grouping of the dose values. |
method |
character string specifying the method to be used to generate a grouping of the dose values. |
breakp |
numeric vector of break points for generating dose intervals that form a grouping. |
display |
logical. If TRUE results are displayed. Otherwise they are not (useful in simulations). |
The functions used the methods df.residual
and residuals
and the 'data'
component of object
(only for determining the number of observations).
The function returns an object of class anova which is displayed using print.anova
.
A clustering technique could be employed to determine the grouping to be used in cases where there are no replicates. There should at most be ceiling(n/2) clusters as otherwise some observations will not be used in the test. At the other end there need to be more clusters than parameters in the model.
Christian Ritz
Neill, J. W. (1988) Testing for lack of fit in nonlinear regression, Ann. Statist., 16, 733–740
See also modelFit
for details on the lack-of-fit test against an ANOVA model.
### Example with 'drc' object ## Lack-of-fit test against ANOVA ryegrass.m1 <-drm(rootl~conc, data = ryegrass, fct = LL.4()) modelFit(ryegrass.m1) ## The same test using 'neill.test' neill.test(ryegrass.m1, ryegrass$conc) ## Generating a grouping neill.test(ryegrass.m1, method="c-finest") neill.test(ryegrass.m1, method="finest") neill.test(ryegrass.m1, method="perc")
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.