* 0.0.21
  - Swictched to check on `env` argument/support in data.table instead of version numbering (https://github.com/Rdatatable/data.table/issues/5466#issuecomment-1272700196)

* 0.0.20
  - `di_ppg` (and hence, `di_iterate`) change:
    + When `reference='hpg'`, the returned `reference_group` column (`di_ppg`) and `ppg_reference_group` column (`di_iterate`) will now show the reference group name (highest performing group) instead of 'hpg'
  - New functions:
    + di_calc_sql
    + di_iterate_sql
    + di_calc_dt
    + di_iterate
  - Scaling DI vignette: updated to reflect demonstrate new functions

* 0.0.19
  - Scaling DI vignette: added explanation of how to use `exclude_scenario_df` parameter for `di_iterate`

* 0.0.18
  - Added `parallel`, `parallel_n_cores`, and `parallel_split_to_disk` parameters to function `di_iterate` for parallel computations, and updated vignette to illustrate the feature
  - Added `suppressMessages` call in key `di_terate` step to reduce dplyr "Joining, by" messages

* 0.0.17
  - Added `custom_reference_group_flag_var` argument to function `di_iterate_on_long` as custom reference groups do not work with the `ppg_reference_groups` and `di_80_index_reference_groups` arguments.

* 0.0.16
  - New function `di_iterate_on_long` allows for DI calculation on long, summarized data sets.
  - New vignette: Disproportionate Impact (DI) Calculations on Long, Summarized Data Sets
  - New data set: ssm_cohort

* 0.0.15
  - student_equity toy data set: added EthnicityFlag_* variables that are more granular than Ethnicity, where students could be assigned to more than 1 group; also modified to force DI in Pacific Islanders
  - Added a Multi-Ethnicity vignette / tutorial
  - Added an Intersectionality vignette / tutorial
  - Changed: prettydoc vignette engine instead of html_document
  - Vignettes: stop creating png files manually

* 0.0.14
  - di_ppg, di_80_index, di_prop_index: return success_needed_not_di, success_needed_full_parity
  - di_iterate: return success_needed_not_di_ppg, success_needed_full_parity_ppg, success_needed_not_di_prop_index, success_needed_full_parity_prop_index, success_needed_not_di_80_index, success_needed_full_parity_80_index

* 0.0.13
  - di_iterate: check if various specified variables are in data
  - di_iterate: fixed the use of exclude_scenario_df (used to returned cases meant to be excluded)
  - Updated tutorial vignette
  - di_ppg bug fix: remove NA when reference group passed and group contains NA (multiple rows returned before)
  - di_80_index bug fix: remove NA when reference group passed and group contains NA (multiple rows returned before)
  - di_80_index: add 'hpg' as the default argument instead of NA so users could understand this better
  - di_iterate: use 'hpg' as default for di_80_index_reference_groups

* 0.0.12
  - Updated Scaling DI vignette
  - di_iterate: check if weight_var exist when specified

* 0.0.11
  - Added `check_valid_reference` argument for `di_iterate`.

* 0.0.10
  - Deprecated `di_ppg_iterate` for `di_iterate`.
  - Updated `di_prop_index` to use a cutoff for determining DI (returning `di_indicator`).
  - Updated `di_ppg`, `di_prop_index`, and `di_80_index` with additional arguments
  - Added a new vignette to describe how to scale DI calculations using `di_iterate`.
  - Fixed `di_ppg_iterate`: since it now uses weight, the calculated references were wrong when a manual reference group was specified.

* 0.0.9
  - Updated `di_ppg_iterate`: remove rows with missing value for success_var, implemented inside `iterate` function within `di_ppg_iterate`.
  - Updated `di_ppg_iterate`: addition of the `weight_var` parameter; if this isn't specified (data set at the individual level), then summarize the input data set to use weights in order to more efficiently calculate DI in multiple iterations.
  - Updated `di_ppg_iterate`: specify columns when using `unnest`.

* 0.0.8
  - Updated `di_ppg_iterate` to filter out scenarios when there are no matched rows when `repeat_by_vars` is used.

* 0.0.7
  - Updated `di_ppg_iterate` to use `slice(i)` instead of manually subsetting rows in order to preserve data frame when `repeat_by_vars` has a single variable
  - Use `tibble` instead of `data_frame` since the latter is deprecated

* 0.0.6
  - Updated `di_ppg_iterate` helper function to include the parameter `repeat_by_vars`.  This allows for repeating DI calculations for every combination of variables specified in `repeat_by_vars`.

* 0.0.5
  - Added `di_ppg_iterate` helper function to iteratively calculate DI across many variables.

* 0.0.4
  - For functions related to the percentage point gap method, add `prop_sub_0` and `prop_sub_1` parameters to account for scenarios where the `proportion` contains 0 or 1 to not lead to a zero MOE when `use_prop_in_moe=TRUE`.  Also, add warning messages when this happens.

* 0.0.3
 - Document definition of `di_indicator` in `di_ppg`
 - Change `pct_hi < reference` to `pct_hi <= reference` in determining DI to be consistent with the methods document from the Chancellor's office