mspathEstimatedCoefficients-class {mspath}R Documentation

mspathEstimatedCoefficients

Description

This class describes estimated coefficients for the multi-state path model. It knows which coefficients are fixed before estimation and which are constrained to be equal. It knows about the errors of the estimates as well. Ordinary users will only care about the arguments to the print function.

Objects from the Class

Create these objects by calling mspathCoefficients(permit, params, offset, baseConstrVec, covVars, constrVec, pathVars, pathConstrVec, fixed=integer(0), n=Inf, var=numeric(0)). var is the estimated variance for all free parameters in this and other coefficients. For other parameters, see slot definitions for details, except that arguments to this function can use more general types like numeric.

Slots

fixed:
Indices, if any, of parameters fixed at the outset ("integer").
n:
Sample size on which estimates are based ("numeric").
foundSE:
"logical" TRUE if the estimate appears to have converged.
var:
"numeric" estimate of the variances of the effective parameters, with 0's for the fixed parameters. Only valid if foundSE == TRUE; otherwise it will be numeric(0). Careful: the variances passed in to the constructor are for all free parameters in all coefficients; this slot has variances for both free and fixed parameters for this coefficient set only.

The remaining items are inheritted from mspathCoefficients:

permit:
"matrix" of "logical" values indicating which entries are permitted. The diagonal is always FALSE.
params:
All the model parameters, both free and fixed, in the canonical order for mspath. ("numeric")
baseConstrVec:
"integer", starting at 1, for constraints on the intercepts.
covVars:
Names of the covariates, if any ("character").
constrVec:
"integer", starting at 1, with constraints on the covariates. Empty vector if no covariates.
pathVars:
Names of path-dependent variables, if any ("character").
pathConstrVec:
"integer", starting at 1, with constraints on the path variables. Empty if no path variables.
map:
map["vname"] gives a matrix of integer's. Each entry that is non-zero indicates that, for variable "vname" in that matrix position, the coefficient comes from the corresponding parameter. Use "intercept" to retrieve the intercepts, and the names of covariates or path-dependent variables to retrieve those values. This slot is a "list".

Extends

Class "mspathCoefficients", directly.

Methods

isAllFixed
signature(object="mspathCoefficients"): usually FALSE, but TRUE if all parameters are fixed.
coef
signature(object="mspathEstimatedCoefficients"): Vector of all coefficients, fixed and free. These are the effective coefficients, meaning that if constraints make two coefficients equal, only a single term will appear for them.
sd
signature(x="mspathCoefficients"), na.rm="ANY": Vector of estimated standard deviations of the results of coef(). The na.rm probably won't work correctly.
print
signature(x="mspathEstimatedCoefficients", coeff, ...): print out the coefficient values. The printout consists of blocks of coefficients related to a particular covariate. Each line indicates the matrix element(s) the coefficient applies to, the index in the parameter list from which the element comes (negative index if fixed), the estimated (or fixed) value of the coefficient, the estimated standard error of the estimate (or NA for fixed values), and various measures of statistical significance.

Parameters that apply to several covariates because of constraints will appear in several places.

The optional argument coeff is a character vector of coefficient names. If provided, only those coefficients will be printed out, in the order given.

... allows one to use further arguments to control the details of the printout; see printCoefmat for details.

Purists will note that neither coeff nor ... is part of the method signature. They are acceptable arguments to the function, however.

Returns invisible(x).

show
signature(object="mspathEstimatedCoefficients"): print the object, returning invisible(NULL).

Note

foundSE will be TRUE if the hessian is positive definite. In that case, var has the diagonal of the inverse of the hessian.

Author(s)

Ross Boylan ross@biostat.ucsf.edu

See Also

mspath gives full information about paramter order and constraints. mspathCoefficients is my parent class. printCoefmat describes additional optional arguments to the print function.

Examples

# 5 x 5, history but no covariates
permissible <- matrix(c( 0, 1, 1, 0, 1,
                         1, 0, 1, 1, 1,
                         0, 0, 0, 0, 1,
                         1, 0, 0, 0, 1,
                         0, 0, 0, 0, 0), ncol=5, byrow=TRUE)
pathConstraint <- c(1, 1, 2, 3, 1, 1, 4, 5, 2, 6,
                   7, 8, 2,  8, 7, 8, 9, 10, 11, 12)
baseConstraint <- c(1, 1, 2, 3, 1, 1, 4, 5, 2, 6)

pathNames <- c("TIS", "TSO")
fixed <- c(3, 10, 11)
var <- seq(from=30, by=5, length.out=15)^2
c <- mspathEstimatedCoefficients(permissible,
                                 params= 101:118,
                                 baseConstrVec=baseConstraint,
                                 pathVars=pathNames,
                                 pathConstrVec = pathConstraint,
                                 fixed=fixed,
                                 n=50,
                                 var=var)

# default show method
c

# optional print arguments
print(c, coeff=c("intercept", "TIS"), digits=2)

[Package mspath version 0.9-9 Index]