Covariates: the fitted terms of a mortality model

covariates() collects the \(X\) terms of a proportional hazards model, \(\log\mu = \log\mu^{\mathrm{ref}} + \sum_j \beta_j X^{(j)}\). Each argument is a pronoun expression, and the coefficients belong to the fit.

disjoint() is the same, and additionally asserts that the terms are mutually exclusive indicators – no individual is in more than one. That is worth stating: where it holds, every off-diagonal entry of the information matrix is zero, and the fit does far less work.

Multiplying by a variable distributes over the list, which is how the common shape \(X^{(k)} = I^{(k)} \varphi\) is written: a set of indicators times a shared age shape. Multiplying two covariate lists gives their Cartesian product.

Usage

covariates(...)

disjoint(...)

is_covariates(x)

is_disjoint(x)

Arguments

...

Pronoun expressions, optionally named. A covariates object is spliced in.

x

An object.

Value

covariates() and disjoint() return a covariates.

is_covariates() returns a scalar logical.

is_disjoint() returns a scalar logical saying whether the terms were asserted mutually exclusive.

Examples

age_shape <- variable(.x)
covariates(.i$sex == "M", .i$sex == "F") * age_shape
#> <covariates: 2 terms>
#>   [1] ((.i$sex == "M") * (.t - .i$birth))
#>   [2] ((.i$sex == "F") * (.t - .i$birth))