Redistribute population over a specified age based on a stable standard fit to the data

Source: R/OPAG.R

This can be used as an external check of population counts in older ages, assuming the stable population standard is representative enough, or it can be used to redistribute population in ages above a specified ages Redistribute_from. This is handy, for instance, for ensuring all censuses extend to a specified maximum age (e.g. 100+) prior to intercensal interpolations. The assumption is that, at least in ages including Age_fit and higher ages, the population should follow a stable pattern proportional to a given survival curve subject to constant growth, r.

OPAG(
  Pop,
  Age_Pop,
  nLx,
  Age_nLx,
  Age_fit = NULL,
  AgeInt_fit = NULL,
  Redistribute_from = max(Age_Pop),
  OAnew = max(Age_nLx),
  method = "mono"
)

Arguments

Pop

numeric vector of population counts
Age_Pop

integer vector of the lower bounds of the population age groups
nLx

numeric vector of stationary population age structure in arbitrary integer age groups
Age_nLx

integer vector of lower bounds of age groups of nLx
Age_fit

integer vector of lower bounds for age groups of Pop_fit
AgeInt_fit

integer vector of widths of age groups of Pop_fit
Redistribute_from

integer lower age bound that forms the cutoff, above which we redistribute counts using the stable standard.
OAnew

integer. Desired open age group in the output (must being element of Age_nLx)
method

character, graduation method used for intermediate graduation. Default "mono". Other reasonable choices include "pclm" or "uniform".

Details

It may be helpful to try more than one fitting possibility, and more than one Redistribute_from cut point, as results may vary.

Redistribute_from can be lower than your current open age group, and OAnew can be higher, as long as it is within the range of Age_nLx. If Age_nLx doesn't go high enough for your needs, you can extrapolate it ahead of time. For this, you'd want the nMx the underlie it, and you can use lt_abridged(), specifying a higher open age, and then extracting nLx again from it.

Examples

# India Males, 1971
Pop            <- smooth_age_5(pop1m_ind,
                         Age = 0:100,
                         method = "Arriaga")
Age_Pop        <- names2age(Pop)
AgeInt_Pop     <- age2int(Age_Pop, OAvalue = 1)

nLx            <- downloadnLx(NULL, "India","male",1971)

#> Downloading nLx data for India, years 1971, gender male
Age_nLx        <- names2age(nLx)
AgeInt_nLx     <- age2int(Age_nLx, OAvalue = 1)

Pop_fit <- OPAG(Pop,
    Age_Pop = Age_Pop,
    nLx = nLx,
    Age_nLx = Age_nLx,
    Age_fit =  c(60,70),
    AgeInt_fit = c(10,10),
    Redistribute_from = 80)

#> 
#> Age_Pop and Age_nLx age intervals are different!

if (FALSE) {
# look at 75+
ind <- Age_Pop >= 75
plot(Age_Pop[ind], Pop[ind])
lines(Age_Pop[ind], Pop_fit$Pop_out[ind], col = "blue")

# relative differences in ages 80+
ind <- Age_Pop >= 80
plot(Age_Pop[ind],  (Pop_fit$Pop_out[ind] - Pop[ind]) / Pop[ind])
}