| print.nhm_score {nhm} | R Documentation |
Print output from a score test of a nhm object
Description
Print output from a score test based on parameters supplied to nhm with score_test=TRUE specified.
Usage
## S3 method for class 'nhm_score'
print(x, which_comp = NULL, ...)
Arguments
x |
An object of class |
which_comp |
Optional vector to specify which of the parameters are to be tested. If omitted, the function will assume all parameters governing non-homogeneity are to be tested. Must be supplied if |
... |
Other parameters to be supplied. Currently ignored. |
Details
The function provides usable output from specifying score_test=TRUE when using nhm. It is most useful to provide a quick(er) test of whether there may be non-homogeneity in a specific model. Note that the model assumes the initial parameters correspond to the constrained maximum likelihood estimate (for instance a model with all the parameters relating to time homogeneity).
The method can be used to compute the local score tests of homogeneity proposed by de Stavola (1988) if type="gompertz" is specified in nhm.
If fisherscore=TRUE in nhm then the expected Fisher information is used. Otherwise, the empirical mean of the squared gradient terms (as used in the BHHH algorithm) is used to estimate the information.
Value
Prints the results of a score test.
Author(s)
Andrew Titman a.titman@lancaster.ac.uk
References
de Stavola BL. Testing Departures from Time Homogeneity in Multistate Markov Processes. Journal of the Royal Statistical Society: Series C (Applied Statistics) 1988. 37. 242-250.