In mathedjoe/animaltracker: Animal Tracker

knitr::opts_chunk$set(echo = TRUE)

Setup

Helper Functions/Libraries

library(animaltracker)
library(dplyr)
library(ggplot2)

Load Correct (manually cleaned) Arizona Data

• Renames 'collar', 'date', 'distancetr', and 'RATE' columns to 'GPS', 'Date', 'Distance', and 'Rate' respectively so names match with candidate data

df_correct <- read.csv('correct.csv', stringsAsFactors = F) %>% 
  dplyr::rename(GPS = collar) %>% 
  dplyr::rename(Date = date) %>% 
  dplyr::rename(Distance = distancetr) %>% 
  dplyr::rename(Rate = RATE) %>% 
  dplyr::mutate(Date = as.Date(Date))

Load Candidate (cleaned by animaltracker) Arizona Data

df_candidate <- read.csv('candidate.csv', stringsAsFactors = F) %>% 
  dplyr::mutate(Date = as.Date(Date))

Calculate Summary Statistics

dfs <- compare_summarise_data(df_correct, 
                       df_candidate, 
                       'gps_arizona.csv', 
                       'date_arizona.csv'
                       )

Summary by GPS Collar

• Close match between latitude, longitude, course
• Candidate has less rows
• Underestimates distance, rate (less variable)
• Overestimates elevation (more variable)

summary(dfs$GPS)

Number of Observations

• Candidate has approximately 90-170 less rows (out of ~6400 - 7500 per GPS collar)

violin_compare(dfs$GPS, GPS, "n", "n_GPS_arizona.png")

Mean Latitude

• Differences are on the order of ~1x10^-5 on average

violin_compare(dfs$GPS, GPS, "meanLat", "meanLat_GPS_arizona.png")

Mean Longitude

• Differences are on the order of ~1x10^-5 - 1x10^-4 on average

violin_compare(dfs$GPS, GPS, "meanLong", "meanLong_GPS_arizona.png")

Mean Distance

• Candidate underestimates distance on average by ~4.5 - 7

violin_compare(dfs$GPS, GPS, "meanDist", "meanDist_GPS_arizona.png")

Mean Course

• Differences are on the order of ~1x10^2 - 1x10^1 on average

violin_compare(dfs$GPS, GPS, "meanCourse", "meanCourse_GPS_arizona.png")

Mean Rate

• Candidate underestimates rate on average by ~0.20 - 0.28

violin_compare(dfs$GPS, GPS, "meanRate", "meanRate_GPS_arizona.png")

Mean Elevation

• Candidate overestimates elevation on average by ~19 - 26

violin_compare(dfs$GPS, GPS, "meanElev", "meanElev_GPS_arizona.png")

Summary by Date

• Close match between latitude, longitude, course
• Candidate has less dates
• Underestimates distance, rate (less variable)
• Overestimates elevation (more variable)

summary(dfs$Date)

Number of Observations

• Candidate tends to have less rows per date (difference between correct and candidate: ~ -1 - 60)

violin_compare(dfs$Date, Date, "n", "n_Date_arizona.png")

Mean Latitude

• Differences are on the order of ~1x10-4 on average

violin_compare(dfs$Date, Date, "meanLat", "meanLat_Date_arizona.png")

Mean Longitude

• Differences are on the order of ~1x10-4 on average

violin_compare(dfs$Date, Date, "meanLong", "meanLong_Date_arizona.png")

Mean Distance

• Candidate underestimates distance on average by ~3.3 - 8.6

violin_compare(dfs$Date, Date, "meanDist", "meanDist_Date_arizona.png")

Mean Course

• Differences are ~ +/- 1 on average**

violin_compare(dfs$Date, Date, "meanCourse", "meanCourse_Date_arizona.png")

Mean Rate

• Candidate underestimates rate on average by ~0.07 - 0.38

violin_compare(dfs$Date, Date, "meanRate", "meanRate_Date_arizona.png")

Mean Elevation

• Candidate overestimates elevation on average by ~34 - 10

violin_compare(dfs$Date, Date, "meanElev", "meanElev_Date_arizona.png")

Summary by GPS and Date

• Close match between latitude, longitude
• Underestimates and overestimates number of rows and course, but difference is a close match on average
• Tends to underestimate distance, rate (less variable)
• Overestimates elevation (variability fluctuates)

daily_summary <- compare_summarise_daily(df_correct, df_candidate, "gps_daily_arizona.csv")

summary(daily_summary)

Number of Observations

• Differs by ~ -5 - 8 rows (correct - candidate) per GPS-date pair

  correct_n <- df_correct %>% 
    dplyr::group_by(GPS, Date) %>% 
    dplyr::summarise(n = n()) %>% 
    dplyr::mutate(Data = "Correct")

  candidate_n <- df_candidate %>% 
    dplyr::group_by(GPS, Date) %>% 
    dplyr::summarise(n = n()) %>% 
    dplyr::mutate(Data = "Candidate")

  plot_data <- dplyr::bind_rows(correct_n, candidate_n)

  ggplot(plot_data, aes(x = Date, y = n, group = Data, color = Data)) +
    geom_line() +
    scale_color_discrete(guide = guide_legend(reverse = T)) +
    facet_wrap(vars(GPS))

Mean Latitude

• Differences are on the order of ~1x10-4 on average

line_compare(df_correct, df_candidate, Latitude, "meanLat_line_arizona.png")

Mean Longitude

• Differences are on the order of ~1x10-4 on average

line_compare(df_correct, df_candidate, Longitude, "meanLong_line_arizona.png")

Mean Distance

• Differs by ~ -4 - 13.6 on average
• Possible influential points:

knitr::kable(daily_summary %>% 
               dplyr::filter((meanDistDiff > quantile(meanDistDiff, 0.75) + 1.5*IQR(meanDistDiff)) |
                             (meanDistDiff < quantile(meanDistDiff, 0.25) - 1.5*IQR(meanDistDiff))) %>%
               dplyr::select(GPS, Date, meanDistDiff))

line_compare(df_correct, df_candidate, Distance, "meanDist_line_arizona.png")

Mean Course

• Differences are ~ +/- 6 on average
• Possible influential points:

knitr::kable(daily_summary %>% 
               dplyr::filter((meanCourseDiff > quantile(meanCourseDiff, 0.75) + 1.5*IQR(meanCourseDiff)) |
                             (meanCourseDiff < quantile(meanCourseDiff, 0.25) - 1.5*IQR(meanCourseDiff))) %>%
               dplyr::select(GPS, Date, meanCourseDiff))

line_compare(df_correct, df_candidate, Course, "meanCourse_line_arizona.png")

Mean Rate

• Differs by ~ -0.16 - 0.64 on average
• Possible influential points:

knitr::kable(daily_summary %>% 
               dplyr::filter((meanRateDiff > quantile(meanRateDiff, 0.75) + 1.5*IQR(meanRateDiff)) |
                             (meanRateDiff < quantile(meanRateDiff, 0.25) - 1.5*IQR(meanRateDiff))) %>%
               dplyr::select(GPS, Date, meanRateDiff))

line_compare(df_correct, df_candidate, Rate, "meanRate_line_arizona.png")

Mean Elevation

• Candidate overestimates elevation on average by ~3 - 43
• Possible influential points:

knitr::kable(daily_summary %>% 
               dplyr::filter((meanElevDiff > quantile(meanElevDiff, 0.75) + 1.5*IQR(meanElevDiff)) |
                             (meanElevDiff < quantile(meanElevDiff, 0.25) - 1.5*IQR(meanElevDiff))) %>%
               dplyr::select(GPS, Date, meanElevDiff))

line_compare(df_correct, df_candidate, Elevation, "meanElev_line_arizona.png")

mathedjoe/animaltracker documentation built on Aug. 12, 2021, 7:46 a.m.