f_prediction_intervall: calculate raw prediction intervalls
In erblast/oetteR: Collection of personal R functions

Description Usage Arguments Details Value See Also Examples

uses an empirical approach to calculate prediction intervalls for each predicted vs. observed value pair. The calculated raw prediction intervalls represent a quite flexible fit which can be used to build a less felxible model of the intervalls.

1
2
3

f_prediction_intervall_raw(df, pred_col, obs_col, intervall = 0.95,
  n_neighbours = 500, rm_outliers = T, bootstrap = F, steps = T,
  verbose = T)

`df`	a dataframe containing predictions and obsvervation pairs
`pred_col`	character vector denoting column with predictions
`obs_col`	character vector denoting column with observed values
`intervall`	double, denoting intervall decision boundary Default: 0.95
`n_neighbours`	integer, denoting the number of neighbouring values to be considered for the intervall calculation, Default: 500
`rm_outliers`	logical, remove outlier based on boxstats definition, Default: T
`bootstrap`	logical, if TRUE intervall decision boundary will be calculated by bootstrapping the population of neighbouring values, if FALSE a normal distribution will be assumed and decision boundary will be calculated based on the mean, Default: F
`steps`	logical, if TRUE predictions will be binned instead of considering the neighbouhood of each point, Default: T
`verbose`	logical

DETAILS

OUTPUT_DESCRIPTION

arrange

## Not run: 

m = lm(price ~ carat + depth, ggplot2::diamonds)

df = tibble( obs = ggplot2::diamonds$price
           , pred = predict(m, newdata = ggplot2::diamonds) ) %>%
   f_prediction_intervall_raw( 'pred','obs', intervall = 0.975) %>%
   f_prediction_intervall_raw( 'pred','obs', intervall = 0.025)

df

ggplot2::ggplot(df) +
  geom_point( aes( x = pred, y = obs), data = dplyr::sample_n(df, 500)
              , alpha = 0.5 ) +
  geom_line( aes(x = pred, y = pred_PI2.5_raw ), size = 1, color = 'darkgreen' ) +
  geom_line( aes(x = pred, y = pred_PI97.5_raw ), size = 1, color = 'darkgreen'  ) +
  geom_line( aes(x = pred, y = pred_mean_raw ), size = 1, color = 'tomato'  )


## End(Not run)