The col values are processed through a median (i.e. low pass) filter
to get .<col>_median, the squared differences between
them .<col>_eps2, and its mean .<col>_sigma.
A col value is considered an outlier if .<col>_eps2 is greater then
.<col>_sigma.
Arguments
- df
a data frame of trajectory data
- col
the variable to filter for outliers
- ksize
the kernel size (must be odd)
- fill
whether to substitute the outlier with the median
- keep
whether to keep intermediate results
Value
a data frame with corrected outliers (if fill is TRUE).
If keep is TRUE the intermediate columns .<col>_median,
.<col>_eps2, .<col>_sigma and .<col>_outlier are included.
Details
This approach is inspired by (copied from ;-) the filter function
in Xavier Olive's Python library
traffic.
See also
Other analysis:
cumulative_distance(),
cumulative_time(),
extract_segment(),
smooth_positions()
Examples
if (FALSE) {
library(readr)
library(dplyr)
library(anytime)
library(trrrj)
library(ggplot2)
ifile <- system.file("extdata", "belevingsvlucht.csv", package = "trrrj")
df <- readr::read_csv(ifile) %>%
mutate(timestamp = anytime::anytime(time, tz = "UTC"))
df1 <- df %>%
filter_outlier(col = baroaltitude, ksize = 17, fill = TRUE, keep = FALSE)
ggplot() +
geom_line(data = df, mapping = aes(x = timestamp, y = baroaltitude), colour = "blue") +
geom_line(data = df1, mapping = aes(x = timestamp, y = baroaltitude), colour = "red")
}