rank_level

Make comparisons between two versions of the same outcome by some grouping variables and rank them by the greatest mean absolute percentage change to the smallest.

rank_level(
  new_data,
  old_data,
  threshold = 0,
  comparison_var,
  id_vars,
  group_vars,
  displace_old_zero = 1e-10
)

Arguments

new_data	data.table() New version of data for comparison.
old_data	data.table() Old version of data for comparison.
threshold	numeric() The threshold must be greater than or equal to zero. It represents the maximum absolute percent difference of comparison_var by group_vars to be included in the analysis. The default is zero.
comparison_var	numeric() Outcome variable to make comparisons over.
id_vars	Vector of variable names that uniquely identify both new_data and old_data.
group_vars	Level to summarize change variables for ranking. group_vars are a subset of id_vars.
displace_old_zero	numeric() Displace zeros in old_data to prevent NaNs in change calculations. Default is 1e-10.

Value

data.table() of ranked merged data.tables, which may be a subset of the input data.tables. Rank number 1 corresponds to the greatest change in comparison_var by group_vars between new_data and old_data, rank number 2 corresponds to the second greatest change in comparison_var by group_vars, etc. NOTE: Ranks may be duplicated by group if they have the same values used for ranking.

Examples

rank_level(new_data,
            old_data,
            threshold = 5,
            comparison_var = "outcome",
            id_vars = c("year", "group"),
            group_vars = "group")
#>    year group new_outcome old_outcome pert_diff max_abs_pert_diff
#> 1:    1     b          12           3       300               300
#> 2:    2     b           8           2       300               300
#> 3:    3     b           4           1       300               300
#> 4:    1     c           0           1      -100               100
#> 5:    2     c           0           2      -100               100
#> 6:    3     c           0           3      -100               100
#> 7:    1     a           2           1       100               100
#> 8:    2     a           4           2       100               100
#> 9:    3     a           6           3       100               100
#>    mean_abs_pert_diff rank
#> 1:                300    1
#> 2:                300    1
#> 3:                300    1
#> 4:                100    2
#> 5:                100    2
#> 6:                100    2
#> 7:                100    2
#> 8:                100    2
#> 9:                100    2