Skip to contents

Function to calculate features matrix

Source: R/get-features.R
get_features.Rd

This function takes input from raw 990 Part IV, VI, and XII, then output a original data frame appended with columns as features needed to calculate governance score.

Usage

get_features(dat)

Arguments

dat

A data.frame containing columns ORG_EIN and correct inputs from 990 Part IV, VI, and XII. See details below. ORG_EIN must be a character string of 9 digits. (This data.frame can contain other columns). See details below for which variable from the 990 efile data must be included to run this function.

Value

A data.frame appended with features needed to run the get_scores function.

Details

990 Efiler data can be downloaded from the NCCS website here. Documentation for the 990 Efiler data can be found here with the data dictionary found here.

These are the variables from each Part/Schedule that are needed as inputs to run this function:

Variables from Part IV:

  • F9_04_AFS_IND_X

  • F9_04_BIZ_TRANSAC_DTK_X

  • F9_04_BIZ_TRANSAC_DTK_FAM_X

  • F9_04_BIZ_TRANSAC_DTK_ENTITY_X

  • F9_04_CONTR_NONCSH_MT_25K_X

  • F9_04_CONTR_ART_HIST_X

Variables from Part VI:

  • F9_06_GVRN_NUM_VOTING_MEMB

  • F9_06_GVRN_NUM_VOTING_MEMB_IND

  • F9_06_GVRN_DTK_FAMBIZ_RELATION_X

  • F9_06_GVRN_DELEGATE_MGMT_DUTY_X

  • F9_06_GVRN_DOC_GVRN_BODY_X

  • F9_06_POLICY_FORM990_GVRN_BODY_X

  • F9_06_POLICY_COI_X

  • F9_06_POLICY_COI_DISCLOSURE_X

  • F9_06_POLICY_COI_MONITOR_X

  • F9_06_POLICY_WHSTLBLWR_X

  • F9_06_POLICY_DOC_RETENTION_X

  • F9_06_POLICY_COMP_PROCESS_CEO_X

  • F9_06_DISCLOSURE_AVBL_OTH_X

  • F9_06_DISCLOSURE_AVBL_OTH_WEB_X

  • F9_06_DISCLOSURE_AVBL_REQUEST_X

  • F9_06_DISCLOSURE_AVBL_OWN_WEB_X

Variables from Part XII:

  • F9_12_FINSTAT_METHOD_ACC_OTH

  • F9_12_FINSTAT_METHOD_ACC_ACCRU_X

  • F9_12_FINSTAT_METHOD_ACC_CASH_X

Variables from Schedule M:

  • SM_01_REVIEW_PROCESS_UNUSUAL_X

Input values may be encoded as "true"/"false", "1"/"0", "X" flags, or blanks; blanks may arrive as NA or "" depending on how the tables were read. Blank or missing yes/no responses are treated as 0 (the governance property is absent), and the result does not depend on the blank encoding. Two derived features can still be NA when their inputs are genuinely unavailable: P6_LINE_1 (share of independent voting members, undefined when the voting-member count is 0 or missing) and P12_LINE_1 (accounting method, when none is reported); get_scores imputes these.

Examples

# get data
data("dat_example", package = "governance")
set.seed(57)
keep_rows <- sample(1:nrow(dat_example), 200)
dat_example <- dat_example[keep_rows, ]

# get features
features_example <- get_features(dat_example)