Visualize variation between the expected & actual percentages

Ideal for a 0/1 dichotomous variable.

plot_expected_proportions(
  df,
  dv,
  ...,
  trunc_length = 100,
  sort_by = c("expected", "actual"),
  threshold = 0.02,
  return_data = FALSE,
  n_field = 9,
  color_over = "navyblue",
  color_under = "red"
)

Arguments

df	data to be analyzed
dv	dependent variable
...	Arguments passed on to refactor_columns split_on variable to split data / group by id_col field to use as ID n_cat for categorical variables, the max number of unique values to keep. This field feeds the forcats::fct_lump(n = ) argument. collapse_by should n_cat collapse by the distance to the grand mean "dv" leaving the extremes as is and grouping factors closer to the grand mean as "other" or should it use size "n" n_quantile for numeric/date fields, the number of quantiles used to split the data into a factor. Fields that have less than this amount will not be changed. n_digits for numeric fields, the number of digits to keep in the breaks ex: [1.2345 to 2.3456] will be [1.23 to 2.34] if n_digits = 2 avg_type mean or median ignore_cols columns to ignore from analysis. Good candidates are fields that have have no duplicate values (primary keys) or fields with a large proportion of null values
trunc_length	length to shorten y-axis labels
sort_by	should data be sorted by expected or actual percentages
threshold	the cut-off (percentage difference) between actual and expected values. This allows the chart to focus on the bigger changes. Use NULL to keep all values
return_data	if TRUE will return a data frame instead of a plot
n_field	the max number of facets to show. The fields are sorted in descending order by those that have the most change (the 'field_delta' column).
color_over	color name/hex code for values that are over-represented
color_under	color name/hex code for values that are under-represented

Examples

# sorted by the expected representation (default)
plot_expected_proportions(
  df = employee_attrition[, 1:5],
  dv = attrition
)

# sorted by the actual representation
plot_expected_proportions(
  df = employee_attrition[, 1:5],
  dv = attrition,
  sort_by = "actual"
)

# you can return the dataframe if you want
plot_expected_proportions(
  df = employee_attrition[, 1:5],
  dv = attrition,
  return_data = TRUE
)
#> # A tibble: 14 x 10
#>    field value     n total expected actual  delta abs_delta field_delta category
#>    <fct> <chr> <int> <int>    <dbl>  <dbl>  <dbl>     <dbl>       <dbl> <chr>   
#>  1 job_~ Dire~    69     5     4.69  2.11   -2.58      2.58       46.8  under   
#>  2 job_~ Rese~    80     2     5.44  0.844  -4.60      4.60       40.4  under   
#>  3 job_~ Sale~    83    33     5.65 13.9     8.28      8.28       40.4  over    
#>  4 job_~ Mana~   102     5     6.94  2.11   -4.83      4.83       40.4  under   
#>  5 job_~ Seni~   106     5     7.21  2.11   -5.10      5.10       46.8  under   
#>  6 job_~ Heal~   131     9     8.91  3.80   -5.11      5.11       40.4  under   
#>  7 job_~ Manu~   145    10     9.86  4.22   -5.64      5.64       40.4  under   
#>  8 job_~ Labo~   259    62    17.6  26.2     8.54      8.54       40.4  over    
#>  9 depa~ Sales   446    92    30.3  38.8     8.48      8.48       18.5  over    
#> 10 job_~ Juni~   534    52    36.3  21.9   -14.4      14.4        46.8  under   
#> 11 job_~ Inte~   543   143    36.9  60.3    23.4      23.4        46.8  over    
#> 12 gend~ Fema~   588    87    40    36.7    -3.29      3.29        6.58 under   
#> 13 gend~ Male    882   150    60    63.3     3.29      3.29        6.58 over    
#> 14 depa~ Rese~   961   133    65.4  56.1    -9.26      9.26       18.5  under   

# an example with more parameters
plot_expected_proportions(
  df = employee_attrition[, 1:5], # data to use
  dv = attrition, # can be a field name or an evaluation
  n_cat = 5, # collapse field values into 5 categories, all else in "Other"
  n_field = 2, # keep the first 2 facets
  threshold = NULL # keep all values
)