cond.glm {cond}

Approximate Conditional Inference for Logistic and Loglinear Models

Description

Performs approximate conditional inference on a scalar parameter of interest in logistic and loglinear models. The output is stored in an object of class cond.

Usage

## S3 method for class 'glm'
cond(object, offset, formula = NULL, family = NULL, 
         data = sys.frame(sys.parent()), pts = 20, 
         n = max(100, 2*pts), tms = 0.6, from = NULL, to = NULL, 
         control = glm.control(...), trace = FALSE, ...)

Arguments

object

a glm object. Families supported are binomial and Poisson with canonical link function.

offset

the covariate occurring in the model formula whose coefficient represents the parameter of interest. May be numerical or a two-level factor. In case of a two-level factor, it must be coded by contrasts and not appear as two dummy variables in the model. Can also be a call to a mathematical function (such as exp, sin, ...) or to a mathematical operator (^, /, ...) applied to a numerical variable. The call must always agree with the label used to identify the corresponding parameter in the glm object passed through the object argument or defined by formula and family. Beware that the label includes the identity function I() if an arithmetic operator was used. Other function types (e.g. factor) and interactions are not admitted.

formula

a formula expression (only if no glm object is defined).

family

a family object defining the variance function (only if no glm object is defined). Families supported are binomial and Poisson with canonical link function.

data

an optional data frame in which to interpret the variables occurring in the formula (only if no glm object is defined).

pts

number of output points (minimum 10) that are calculated exactly. The default is 20.

n

approximate number of output points (minimum 50) produced by the spline interpolation. The default is the maximum between 100 and twice pts.

tms

defines the range MLE +/- tms * s.e. where cubic spline interpolation is replaced by polynomial interpolation. The default is 0.6.

from

starting value of the sequence that contains the values of the parameter of interest for which output points are calculated exactly. The default is MLE - 3.5 * s.e.

to

ending value of the sequence that contains the values of the parameter of interest for which output points are calculated exactly. The default is MLE + 3.5 * s.e.

control

a list of iteration and algorithmic constants that controls the GLM fit. See \ glm.control for their names and default values.

trace

if TRUE, iteration numbers will be printed.

...

additional arguments, such as subset etc., used by the glm fitting routine if the glm object is defined through formula and family. See glm for their definition and use. The arguments weights, offset and contrasts are not admitted. The returned value is an object of class cond; see cond.object for details.

Details

This function is a method for the generic function cond for class glm. It can be invoked by calling cond for an object of the appropriate class, or directly by calling cond.glm regardless of the class of the object. cond.glm has also to be used if the glm object is not provided throught the object argument but specified by formula and family.

The function cond.glm implements several small sample asymptotic methods for approximate conditional inference in logistic and loglinear models. Approximations for both the conditional log likelihood function and conditional tail area probabilities are available (see cond.object for details). Attention is restricted to a scalar parameter of interest. The associated covariate can be either numerical or a two-level factor.

Approximate conditional inference is performed by either updating a fitted generalized linear model or defining the model formula and family. All approximations are calculated exactly for pts equally spaced points ranging from from to to. A cubic spline interpolation is used to extend them over the whole interval of interest, except for the range of values defined by MLE +/- tms * s.e. where the spline interpolation is replaced by a higher order polynomial interpolation. This is done in order to avoid numerical instabilities which are likely to occur for values of the parameter of interest close to the MLE. Results are stored in an object of class cond. Method functions like print, summary and plot can be used to examine the output or represent it graphically. Components can be extracted using coef, formula and family.

Main references for the methods considered are the papers by Pierce and Peters (1992) and Davison (1988). More details on the implementation are given in Brazzale (1999, 2000).

Value

The returned value is an object of class cond; see cond.object for details.

Note

In rare occasions, cond.glm dumps because of non-convergence of the function glm which is used to refit the model for a fixed value of the parameter of interest. This happens for instance if this value is too extreme. The arguments from and to may then be used to limit the default range of MLE +/- 3.5 * s.e. A further possibility is to fine-tuning the constants (number of iterations, convergence threshold) that control the GLM fit through the control argument.

cond.glm may also dump if the estimate of the parameter of interest is large (tipically > 400) in absolute value. This may be avoided by reparametrizing the model.

The output of cond.glm may be unreliable if part of the data have a degenerate distribution. For example take the fungal infections treatment data contained in the fungal data frame. Of the five 2\times 2 contingency tables, two (the first and the third) are degenerate. As they make no contribution to the exact conditional likelihood, they should be omitted from the approximate conditional fit.

References

Brazzale, A. R. (1999) Approximate conditional inference for logistic and loglinear models. J. Comput. Graph. Statist., 8, 1999, 653–661.

Brazzale, A. R. (2000) Practical Small-Sample Parametric Inference. Ph.D. Thesis N. 2230, Department of Mathematics, Swiss Federal Institute of Technology Lausanne.

Davison, A. C. (1988) Approximate conditional inference in generalized linear models. J. R. Statist. Soc. B, 50, 445–461.

Pierce, D. A. and Peters, D. (1992) Practical use of higher order asymptotics for multiparameter exponential families (with Discussion). J. R. Statist. Soc. B, 54, 701–737.

See Also

cond.object, summary.cond, plot.cond, glm

Examples

## Crying Babies Data
data(babies)
babies.glm <- glm(formula = cbind(r1, r2) ~ day + lull - 1, 
                  family = binomial, data = babies)
babies.cond <- cond(object = babies.glm, offset = lullyes)
babies.cond
##
## If one wishes to avoid the generalized linear model fit:
babies.cond <- cond.glm(formula = cbind(r1, r2) ~ day + lull - 1, 
                        family = binomial, data = babies, offset = lullyes)
babies.cond

## Urine Data 
## (function call as offset variable) 
data(urine)
urine.glm <- glm(r ~ gravity + ph + osmo + conduct + urea + log(calc), 
                 family = binomial, data = urine)
labels(coef(urine.glm))
urine.cond <- cond(urine.glm, log(calc))
##
## (large estimate of regression coefficient)
urine.glm <- glm(r ~ gravity + ph + osmo + conduct + urea + calc, 
                 family = binomial, data = urine)
coef(urine.glm)
urine.glm <- glm(r ~ I(gravity * 100) + ph + osmo + conduct + urea + calc, 
                 family = binomial, data = urine)
coef(urine.glm)
urine.cond <- cond(urine.glm, I(gravity * 100))

## Fungal Infections Treatment Data (numerical instabilities around the
##                                   MLE)
## (full data analysis)
data(fungal)
fungal.glm <- glm(cbind(success, failure) ~ center + group - 1, 
                  family = binomial, data = fungal, 
                  control = glm.control(maxit = 50, epsilon = 1e-005))
fungal.cond <- cond(fungal.glm, groupT)
plot(fungal.cond, which = 2)
## (partial data analysis)
fungal.glm <- glm(cbind(success, failure) ~ center + group - 1, 
                  family = binomial, data = fungal, subset = -c(1,2,5,6), 
                  control = glm.control(maxit = 50, epsilon = 1e-005))
fungal.cond <- cond(fungal.glm, groupT)
plot(fungal.cond, which = 2)
## (Tables 1 and 3 are omitted).