Skip to contents

Sample size for a subpopulation (domain) within a larger frame

Source: R/subpop_design.R
morie_sample_size_domain.Rd

Answers the planning question "how large must the whole sample be so that the subgroup is itself large enough?". The subgroup requirement is computed first, exactly as in morie_sample_size_proportion; the overall requirement is then that subgroup size divided by the subgroup's share of the frame, further divided by the coverage rate if the frame under-identifies the group.

Usage

morie_sample_size_domain(
  p,
  moe,
  domain_prevalence,
  conf = 0.95,
  N_domain = Inf,
  deff = 1,
  response_rate = 1,
  coverage = 1
)

Arguments

p

Anticipated proportion within the subgroup.

moe

Target margin of error for the subgroup estimate.

domain_prevalence

Share of the frame belonging to the subgroup.

conf

Confidence level. Default 0.95.

N_domain

Subgroup population size. Default Inf.

deff

Design effect within the subgroup. Default 1.

response_rate

Expected response rate. Default 1.

coverage

Proportion of subgroup members correctly identified as such by the frame. Default 1.

Value

A list with n_domain (completed subgroup interviews needed), n_domain_invite, n_overall (units that must be selected from the full frame under proportionate sampling), and the inputs.

Details

Two facts drive the result and both are easy to lose. A subgroup forming a small share of the population requires a very large general sample for even a modest subgroup target, which is why general-population surveys are usually unable to report on small subpopulations without oversampling. And if the frame identifies only a fraction of the group, that fraction divides the achievable subgroup size a second time.

See also

morie_oversample_factor, morie_screen_design

Examples

# Subgroup is 5% of the frame; want 5 points on a 50% attribute
morie_sample_size_domain(0.5, 0.05, domain_prevalence = 0.05)
#> $n_domain
#> [1] 384.1459
#> 
#> $n_domain_invite
#> [1] 384.1459
#> 
#> $n_overall
#> [1] 7682.918
#> 
#> $domain_prevalence
#> [1] 0.05
#> 
#> $coverage
#> [1] 1
#> 
#> $deff
#> [1] 1
#> 
#> $response_rate
#> [1] 1
#> 
#> $moe
#> [1] 0.05
#> 
#> $conf
#> [1] 0.95
#>