Introduction to dplyr, using SAO data

df <- read.csv('State_s_Attorney_Felony_Cases_-_Disposition_Outcomes_By_Offense_Type_and_Defendant_Race.csv')
kable(head(df))

unique_dispositions <- unique(df$DISPOSITION)
length(unique_dispositions)

## [1] 33

unique_dispositions[grepl('^Plea', unique_dispositions)]

## [1] Plea Of Guilty                   Plea of Guilty - Amended Charge
## [3] Plea of Guilty - Lesser Included Plea of Guilty But Mentally Ill
## 33 Levels:  BFW Case Dismissed Charge Rejected ... WOWI

prep_disaggregated <- function(df) {
  df <- df %>%
    # Exclude blank dispositions, there are a few
    filter(DISPOSITION != '') %>%
    # Keep year, disposition, and offense type
    group_by(YEAR, DISPOSITION, OFFENSE.TYPE) %>%
    # Sum up the total number of defendants
    summarise(total_defendants=sum(DEFENDANT.COUNT)) %>%
    # Add an extra column to determine whether the particular disposition was a guilty plea. There
    # are 4 different dispositions for guilty pleas, which we can identify with the '^Plea' regular
    # expression
    mutate(
      pled_guilty = grepl('^Plea', DISPOSITION)
    )

  return(df)
}
disagg <- prep_disaggregated(df)
kable(head(disagg))

narcotics_vs_other_plea_rates <- function(df) {
  # Expects to have a "pled_guilty" column already present
  plea_rate_by_narco <- df %>%
    # Add a boolean column to indicate whether the offense was a narcotics cases or not
    mutate(narcotics_case = OFFENSE.TYPE == 'Narcotics') %>%
    # Now group by that new column - dplyr is flexible enough to use "new" column in expressions that
    # follow
    group_by(narcotics_case) %>%
    # Now that we're grouping by the `narcotics_case` boolean, sum up the total number of cases, the number
    # of cases that were pled guilt (as indicated by the `pled_guilty` field), the the percentage of cases
    # pled guilty. Note that columns used in `summarise` can reference one another - e.g. pct_pled_guilty
    # is a function of two other columns in that same `summarise` invocation.
    summarise(
      total_cases = sum(total_defendants),
      num_pled_guilty = sum(total_defendants[pled_guilty]),
      pct_pled_guilty= num_pled_guilty / total_cases
    )
  return(plea_rate_by_narco)
}
plea_rates_by_narco <- narcotics_vs_other_plea_rates(disagg)
kable(plea_rates_by_narco)

# Check total number of narcotics cases
only_narcotics_cases <- disagg %>%
  filter(OFFENSE.TYPE == 'Narcotics')
sum(only_narcotics_cases$total_defendants)

## [1] 101924

# Check total number of _non_ narcotics cases
non_narcotics_cases <- disagg %>%
  filter(OFFENSE.TYPE != 'Narcotics')
sum(non_narcotics_cases$total_defendants)

## [1] 125918

# Check total number of pled-out narcotics cases
pled_out_narcotics_cases <- disagg %>%
  filter(OFFENSE.TYPE == 'Narcotics') %>%
  filter(pled_guilty)
sum(pled_out_narcotics_cases$total_defendants)

## [1] 43866

# Check total number of pled-out _non_ narcotics cases
pled_out_non_narcotics_cases <- disagg %>%
  filter(OFFENSE.TYPE != 'Narcotics') %>%
  filter(pled_guilty)
sum(pled_out_non_narcotics_cases$total_defendants)

## [1] 95295

most_common_dispos <- function(df) {
  num_all_dispos <- sum(df$total_defendants)
  most_common <- df %>%
    # We're interested in individual dispositions, so group by them here
    group_by(DISPOSITION) %>%
    # Add columns for the total number of cases per disposition, and the % of all cases
    summarise(
      total_dispos = sum(total_defendants),
      pct_of_total = total_dispos / num_all_dispos
    ) %>%
    # Arrange the rows by our newly added `total_dispos` column. Using `desc` orders the rows in descending
    # order, using the `total_dispos` column.
    arrange(desc(total_dispos))

  return(most_common)
}
most_common_narcotics_dispos <- most_common_dispos(only_narcotics_cases)
kable(head(most_common_narcotics_dispos))