poputils (version 0.4.1)

ex_to_lifetab_brass: Derive Life Tables that Match Life Expectancies, using a Brass Logit Model

Description

Turn life expectancies at birth into full life tables, using the Brass logit model. The method is simple and is designed for simulations or for settings with little or no data on age-specific mortality rates. In settings where data on age-specific mortality is available, other methods might be more appropriate.

Usage

ex_to_lifetab_brass(
  target,
  standard,
  infant = c("constant", "linear", "CD", "AK"),
  child = c("constant", "linear", "CD"),
  closed = c("constant", "linear"),
  open = "constant",
  radix = 1e+05,
  suffix = NULL
)

Value

A data frame containing one or more life tables.

Arguments

target

A data frame containing a variable called "ex", and possibly others. See Details.

standard

A data frame containing variables called age and lx, and possibly others. See details.

infant, child, closed, open

Methods used to calculate life expectancy. See lifetab() for details.

radix

Initial population for the lx column in the derived life table(s). Default is 100000.

suffix

Optional suffix added to life table columns.

Method

The method implemented by ex_to_lifetab_brass() is based on the observation that, if populations A and B are demographically similar, then, in many cases,

$$\text{logit}(l_x^{\text{B}}) \approx \alpha + \beta \text{logit}(l_x^{\text{A}})$$

where \(l_x\) is the "survivorship probability" quantity from a life table. When populations are similar, \(beta\) is often close to 1.

Given (i) target life expectancy, (ii) a set of \(l_x^{\text{A}}\)), (referred to as a "standard"), and (iii) a value for \(\beta\), ex_to_lifetab_brass() finds a value for \(\alpha\) that yields a set of \(l_x^{\text{B}}\)) with the required life expectancy.

<code>target</code> argument

target is a data frame specifying life expectancies for each population being modelled, and, possibly, inputs to the calculations, and index variables. Values in target are not age-specific.

• A variable called "ex", with life expectancy at birth must be included in target.

• A variable called "beta" with values for beta can be included in target. This variable can be an rvec. If no "beta" variable is included in target, then ex_to_lifetab_brass() assumes that \(beta \equiv 1\).

• A variable called "sex". If the infant argument to ex_to_lifetab_brass() is is "CD" or "AK", or if the child argument is "CD", target must include a "sex" variable, and the labels for this variable must be interpretable by function [format_sex()]. Otherwise, the "sex"` variable is optional, and there is no restriction on labels.

• Other variables used to distinguish between life expectancies, such as time, region, or model variant.

<code>standard</code> argument

standard is a data frame specifying the \(l_x\) to be used with each life expectancy in ex, and, optionally, values the average age person-years lived by people who die in each group, \(_na_x\). Values in standard are age-specific.

• A variable called "age", with labels that can be parsed by reformat_age().

• A variable called "lx". Internally each set of \(l_x\) is are standardized so that the value for age 0 equals 1. Within each set, values must be non-increasing. Cannot be an rvec.

• Additional variables used to match rows in standard to rows in target.

Internally, standard is merged with target using a left join from target, on any variables that target and standard have in common.

References

Brass W, Coale AJ. 1968. “Methods of analysis and estimation,” in Brass, W, Coale AJ, Demeny P, Heisel DF, et al. (eds). The Demography of Tropical Africa. Princeton NJ: Princeton University Press, pp. 88–139.

Moultrie TA, Timæus IM. 2013. Introduction to Model Life Tables. In Moultrie T, Dorrington R, Hill A, Hill K, Timæus I, Zaba B. (eds). Tools for Demographic Estimation. Paris: International Union for the Scientific Study of Population. online version.

See Also

• logit(), invlogit() Logit function

• lifeexp() Calculate life expectancy from detailed inputs

Examples

## create new life tables based on level-1
## 'West' model life tables, but with lower
## life expectancy

library(dplyr, warn.conflicts = FALSE)

target <- data.frame(sex = c("Female", "Male"), 
                     ex = c(17.5, 15.6))

standard <- west_lifetab |>
    filter(level == 1) |>
    select(sex, age, lx)
    
ex_to_lifetab_brass(target = target,
                    standard = standard,
                    infant = "CD",
                    child = "CD")