mselect.RdModel selection by comparison of different models using the following criteria: the log likelihood value, Akaike's information criterion (AIC), the estimated residual standard error or the p-value from a lack-of-fit test.
mselect(object, fctList = NULL, nested = FALSE, sorted = c("IC", "Res var", "Lack of fit", "no"), linreg = FALSE, icfct = AIC)
| object | an object of class 'drc'. |
|---|---|
| fctList | a list of dose-response functions to be compared. |
| nested | logical. TRUE results in F tests between adjacent models (in 'fctList'). Only sensible for nested models. |
| sorted | character string determining according to which criterion the model fits are ranked. |
| linreg | logical indicating whether or not additionally polynomial regression models (linear, quadratic, and cubic models) should be fitted (they could be useful for a kind of informal lack-of-test consideration for the models specified, capturing unexpected departures). |
| icfct | function for supplying the information criterion to be used. |
For Akaike's information criterion and the residual standard error: the smaller the better and for lack-of-fit test (against a one-way ANOVA model): the larger (the p-value) the better. Note that the residual standard error is only available for continuous dose-response data.
Log likelihood values cannot be used for comparison unless the models are nested.
A matrix with one row for each model and one column for each criterion.
### Example with continuous/quantitative data ## Fitting initial four-parameter log-logistic model ryegrass.m1 <- drm(rootl ~ conc, data = ryegrass, fct = LL.4()) ## Model selection mselect(ryegrass.m1, list(LL.3(), LL.5(), W1.3(), W1.4(), W2.4(), baro5()))#> logLik IC Lack of fit Res var #> W2.4 -15.91352 41.82703 0 0.2646283 #> LL.4 -16.15514 42.31029 0 0.2700107 #> baro5 -15.86422 43.72844 0 0.2774141 #> LL.5 -15.87828 43.75656 0 0.2777393 #> W1.4 -17.46720 44.93439 0 0.3012075 #> LL.3 -18.60413 45.20827 0 0.3153724 #> W1.3 -22.22047 52.44094 0 0.4262881## Model selection including linear, quadratic, and cubic regression models mselect(ryegrass.m1, list(LL.3(), LL.5(), W1.3(), W1.4(), W2.4(), baro5()), linreg = TRUE)#> logLik IC Lack of fit Res var #> W2.4 -15.91352 41.82703 0 0.2646283 #> LL.4 -16.15514 42.31029 0 0.2700107 #> baro5 -15.86422 43.72844 0 0.2774141 #> LL.5 -15.87828 43.75656 0 0.2777393 #> W1.4 -17.46720 44.93439 0 0.3012075 #> LL.3 -18.60413 45.20827 0 0.3153724 #> W1.3 -22.22047 52.44094 0 0.4262881 #> Cubic -25.53428 61.06856 NA 0.5899609 #> Quad -35.11558 78.23116 NA 1.2485122 #> Lin -50.47554 106.95109 NA 4.2863247#> logLik IC Lack of fit Res var Nested F test #> LL.4 -16.15514 42.31029 0.8664830 0.2700107 NA #> LL.5 -15.87828 43.75656 0.8538476 0.2777393 0.5134602### Example with quantal data ## Fitting initial two-parameter log-logistic model earthworms.m1 <- drm(number/total~dose, weights=total, data = earthworms, fct = LL.2(), type = "binomial") ## Comparing 4 models mselect(earthworms.m1, list(W1.2(), W2.2(), LL.3()))#> logLik IC Lack of fit #> LL.3 -36.15518 78.31036 0 #> LL.2 -347.55013 699.10026 0 #> W2.2 -347.63989 699.27978 0 #> W1.2 -348.82481 701.64963 0