Skip to contents

Calculate maximum error across weighted variables

calc_max_error.Rd

Calculates and returns the maximum absolute percent error among weighted variables comparing survey incidence to targets. Use to identify the variable with the largest fit discrepancy after weighting.

Usage

calc_max_error(
  seed_incidence_table,
  target_table,
  weights,
  weight_col = "final_weight"
)

Arguments

seed_incidence_table

data.table with required columns:

  • hh_id_dow — household-day identifier

  • p_* — person-level variables (prefix p_)

  • h_* — household-level variables (prefix h_) Rows: one per household-day. Keys: (hh_id_dow). Modified by reference: no (returns copy).

target_table

data.table with required columns:

  • hh_id_dow — household-day identifier

  • p_* — person-level target variables (prefix p_)

  • h_* — household-level target variables (prefix h_) Rows: one per household-day. Keys: (hh_id_dow). Modified by reference: no (returns copy).

weights

data.table with required columns:

  • hh_id_dow — household-day identifier

  • final_weight — weight value (or as specified by weight_col) Rows: one per household-day. Keys: (hh_id_dow). Modified by reference: no (returns copy).

weight_col

character(1). Name of the column in weights containing the weight values. Default: 'final_weight'.

Value

named numeric(1). Maximum absolute percent error, with variable name as name.

Details

  • Checks that household IDs are consistent and ordered between tables.

  • Computes weighted sums for all target columns (prefix p_ or h_).

  • Calculates absolute percent error for each variable.

  • Returns the maximum error and its variable name.

  • Returns a named numeric vector (variable name as name).

Settings

None.

See also

calc_weight_mape calc_weight_fit scripts/weighting/fit_analysis.R

Other survey analysis: calc_survey_ci(), calc_weight_fit()

Examples

## Not run:
seed_incidence_table <- data.table(hh_id_dow = 1:3, p_var1 = c(1,2,3), h_var2 = c(4,5,6))
target_table <- data.table(hh_id_dow = 1:3, p_var1 = c(1,2,3), h_var2 = c(4,5,6))
weights <- data.table(hh_id_dow = 1:3, final_weight = c(1,1,1))
calc_max_error(seed_incidence_table, target_table, weights)
#> Error in weighted_sums - target_vec: non-conformable arrays
## End(Not run)