Skip to contents

Classify a window summary on the SPI and the NPAFP rate

Source: R/spi_rolling.R
spi_window_classify.Rd

Sets the conventional non-polio AFP rate against its target and the SPI against its threshold, for a spi_window_summary() result.

The rate has the observed count over the window as numerator and pop_u15 / rate_multiplier as denominator. It is not annualised by default, so it is a yearly rate when the window covers 12 months. With annualise = TRUE the same ratio is multiplied by months_per_year and divided by months. A unit meets the rate target when the rate is at or above rate_target.

A unit is flagged by the SPI, meaning low relative AFP reporting, when spi_median is below index_threshold and the upper bound spi_q95 is below 1, so the whole 90% interval sits below the national level. A unit whose expected_total is below min_expected, or whose SPI is missing, carries too little information and is left unclassified (NA).

This relates to spi_compare_npafp() as follows. With spi_rule = "interval", spi_compare_npafp() flags the SPI by the same rule, but it takes a district-year spi_index() or spi_simple() result, computes the rate as count / population x multiplier from a population table, classifies every district-year without a min_expected rule, uses the labels "Neither below", "SPI below threshold only", "NPAFP below target only" and "Both below", and adds agreement statistics. spi_window_classify() works on one rolling window, leaves low-information units unclassified, and returns the classification column by column.

Usage

spi_window_classify(
  summary,
  rate_target = 3,
  index_threshold = 0.8,
  min_expected = 1,
  rate_multiplier = 1e+05,
  annualise = FALSE,
  months_per_year = 12,
  agreement_levels = c("both_pass", "both_flag", "spi_only", "rate_only"),
  focus_levels = c("focus", "other")
)

Arguments

summary

A spi_window_summary() result.

rate_target

Numeric. NPAFP rate target per rate_multiplier. Default: 3.

index_threshold

Numeric. SPI below which a unit can be flagged. Default: 0.8.

min_expected

Numeric. Smallest posterior median expected count at which a unit is classified. Default: 1.

rate_multiplier

Numeric. Population scale of the rate. Default: 100000.

annualise

Logical. Scale the rate to months_per_year. Default: FALSE.

months_per_year

Numeric. Months the rate is annualised to when annualise = TRUE. Default: 12.

agreement_levels

Character, length 4. Labels for: both pass, both flag, rate passes and SPI flags, rate flags and SPI passes. Default: c("both_pass", "both_flag", "spi_only", "rate_only").

focus_levels

Character, length 2. Labels for units that meet the rate target but are flagged by the SPI, and for every other classified unit. Default: c("focus", "other").

Value

summary with npafp_rate, readable, rate_pass, ci_below, index_pass, agreement (factor) and focus_status (factor) added.

See also

Other spi rolling window functions: spi_rolling_index(), spi_window_summary()

Examples

summary <- tibble::tibble(
  observed = c(3, 3), pop_u15 = 1e5, months = 12L,
  expected_total = 5, spi_median = c(0.9, 0.7), spi_q95 = c(1.1, 0.9)
)
spi_window_classify(summary)
#> # A tibble: 2 × 13
#>   observed pop_u15 months expected_total spi_median spi_q95 npafp_rate readable
#>      <dbl>   <dbl>  <int>          <dbl>      <dbl>   <dbl>      <dbl> <lgl>   
#> 1        3  100000     12              5        0.9     1.1          3 TRUE    
#> 2        3  100000     12              5        0.7     0.9          3 TRUE    
#> # ℹ 5 more variables: rate_pass <lgl>, ci_below <lgl>, index_pass <lgl>,
#> #   agreement <fct>, focus_status <fct>