REMLRT.asreml {asremlPlus}

Performs a REML ratio test to compare two models.

Description

Extracts the REML log likelihood and the number of variance parameters from two asreml objects. It assumes that the first asreml object corresponds to the null hypothesis and the second asreml object to the alternative hypothesis for the test being conducted. That is, the second asreml object is the result of fitting a model that is a reduced version of the model for the first object. In the case where the reduced model is obtained by setting positively-constrained variance parameters in the full model to zero, the positive.zero argument should be set to TRUE so that the p-value is computed using a mixture of chi-square distributions as described in Self and Liang (1987).

The function checks that the models do not differ in either their fixed or sparse models.

Usage

## S3 method for class 'asreml'
REMLRT(h0.asreml.obj, h1.asreml.obj, 
       positive.zero = FALSE, bound.test.parameters = "none", 
       DF = NULL, bound.exclusions = c("F","B","S","C"), ...)

Arguments

h0.asreml.obj

asreml object containing the fit under the model for the null hypothesis.

h1.asreml.obj

asreml object containing the fit under the model for the alternative hypothesis.

positive.zero

Indicates whether the hypothesized values for the variance components being tested are on the boundary of the parameter space. For example, this is true for positively-constrained variance components that, under the reduced model, are zero. This argument does not need to be set if bound.test.parameters is set.

bound.test.parameters

Indicates whether for the variance components being tested, at least some of the hypothesized values are on the boundary of the parameter space. The possibilities are "none", "onlybound" and "one-and-one". The default is "none", although if it is set to "none" and positive.zero is TRUE then bound.test.parameters is taken to be "onlybound". When bound.test.parameters is set to "one-and-one", it signifies that there are two parameters being tested, one of which is bound and the other is not. For example, the latter is true for testing a covariance and a positively-constrained variance component that, under the reduced model, are zero.

DF

A numeric giving the difference between the two models in the number of variance parameters whose estimates are not of the type specified in bound.exclusions. If NULL then this is determined from the information in full.asreml.obj and reduced.asreml.obj.

bound.exclusions

A character specifying one or more bound (constraint) codes that will result in a variance parameter being excluded from the count of estimated variance parameters. If set to NULL then none will be excluded.

...

Provision for passing arguments to functions called internally - not used at present.

Value

A data.frame containing the log of the likelihood ratio, its degrees of freedom, its p-value and the number of bound parameters in each of the two models being compared.

Note

If DF is not NULL, the supplied value is used. Otherwise DF is determined from the information in h1.asreml.obj and h0.asreml.obj. In this case, the degrees of freedom for the test are computed as the difference between the two models in the number of variance parameters whose estimates do not have a code for bound specified in bound.exclusions.

If ASReml-R version 4 is being used then the codes specified in bound.exclusions are not restricted to a subset of the default codes, but a warning is issued if a code other than these is specified. For ASReml-R version 3, only a subset of the default codes are allowed: F (Fixed), B (Boundary), C (Constrained) and S (Singular).

The test statistic is calculated as 2(log(REML)_1 - log(REML)_0).

This procedure is only appropriate when the null hypothesis is that (i) all parameters are on the boundary of the parameter space (ii) all parameters are in the interior of the parameter space, or (iii) there are two parameters, one of which is on the boundary and the other is not. Other cases have been discussed by Self and Liang (1987), but are not implemented here.

Author(s)

Chris Brien

References

Self, S.G., and Liang, K-Y. (1987) Asymptotic Properties of Maximum Likelihood Estimators and Likelihood Ratio Tests Under Nonstandard Conditions. Journal of the American Statistical Association, 82, 605-10.

See Also

infoCriteria.asreml, testranfix.asrtests

Examples

## Not run: 
    REMLRT(ICV.max, ICV.red, bound.test.parameters = "onlybound")

## End(Not run)