Likelihood ratio test between two or more fitted multi-state models
Arguments
- ...
Two or more fitted multi-state models, as returned by
msm, ordered by increasing numbers of parameters.
Value
A matrix with three columns, giving the likelihood ratio statistic, difference in degrees of freedom and the chi-squared p-value for a comparison of the first model supplied with each subsequent model.
Warning
The comparison between models will only be valid if they are fitted to the same dataset. This may be a problem if there are missing values and R's default of 'na.action = na.omit' is used.
The likelihood ratio statistic only has the indicated chi-squared
distribution if the models are nested. An alternative for comparing
non-nested models is Akaike's information criterion. This can be computed
for one or more fitted msm models x,y,... using
AIC(x,y,...).