VGAM (version 0.8-2)

fff: F Distribution Family Function

Description

Maximum likelihood estimation of the (2-parameter) F distribution.

Usage

fff(link="loge", earg=list(), idf1=NULL, idf2=NULL, nsimEIM=100,
    method.init=1, zero=NULL)

Arguments

link

Parameter link function for both parameters. See Links for more choices. The default keeps the parameters positive.

earg

List. Extra argument for the link. See earg in Links for general information.

idf1, idf2

Numeric and positive. Initial value for the parameters. The default is to choose each value internally.

nsimEIM

See CommonVGAMffArguments for more information.

method.init

Initialization method. Either the value 1 or 2. If both fail try setting values for idf1 and idf2.

zero

An integer-valued vector specifying which linear/additive predictors are modelled as intercepts only. The value must be from the set {1,2}, corresponding respectively to $df1$ and $df2$. By default all linear/additive predictors are modelled as

Value

• An object of class "vglmff" (see vglmff-class). The object is used by modelling functions such as vglm and vgam.

Warning

Numerical problems will occur when the estimates of the parameters are too low or too high.

Details

The F distribution is named after Fisher and has a density function that has two parameters, called df1 and df2 here. This function treats these degrees of freedom as positive reals rather than integers. The mean of the distribution is $df2/(df2-2)$ provided $df2>2$, and its variance is $2 df2^2 (df1+df2-2)/(df1 (df2-2)^2 (df2-4))$ provided $df2>4$. The estimated mean is returned as the fitted values. Although the F distribution can be defined to accommodate a non-centrality parameter ncp, it is assumed zero here. Actually it shouldn't be too difficult to handle any known ncp; something to do in the short future.

References

Evans, M., Hastings, N. and Peacock, B. (2000) Statistical Distributions, New York: Wiley-Interscience, Third edition.

See Also

FDist.

Examples

x = runif(n <- 2000)
df1 = exp(2+0.5*x)
df2 = exp(2-0.5*x)
y = rf(n, df1, df2)
fit = vglm(y  ~ x, fff, trace=TRUE)
coef(fit, matrix=TRUE)