0

I'm working with a large dataset of different variables collected during the dives of elephant seals. I would like to analyze my data on a fine-scale (20 second intervals). I want to bin my data into 20 second intervals, basically I just want to get the mean for every 20 seconds, so I can run more analysis on these intervals of data. However, I need to group my data by dive # so that I'm not binning information from separate dives.

There are three methods I've tried so far:

  • period.apply() but I cannot group with this function.
  • split() to subset my data by dive #, but can't seem to find a way to then calculate the mean of different columns over 20 second intervals within these subsets.
  • openair package, using timeaverage() but continue to get an error (see code below).

Below is what the data looks like, and the code I've tried. I would like the means of Depth, MSA, rate_s, and HR for each 20 second window - grouped by diveNum and ~ideally~ also D_phase.

> head(seal_dives)
             datetime   seal_ID  Depth    MSA        D_phase diveNum rate_s     HR
1 2018-04-06 14:47:51  Congaree  4.5    0.20154042       D       1     NA     115.3846
2 2018-04-06 14:47:51  Congaree  4.5    0.20154042       D       1     NA     117.6471
3 2018-04-06 14:47:52  Congaree  4.5    0.11496760       D       1     NA     115.3846
4 2018-04-06 14:47:52  Congaree  4.5    0.11496760       D       1     NA     122.4490
5 2018-04-06 14:47:53  Congaree  4.5    0.05935992       D       1     NA     113.2075
6 2018-04-06 14:47:53  Congaree  4.5    0.05935992       D       1     NA     113.2075

#openair package using timeaverage, results in error message
> library(openair)
> seal_20<-timeAverage(
   seal_dives,
   avg.time = "20 sec",
   data.thresh = 0,
   statistic = "mean",
   type = c("diveNum","D_phase"),
   percentile = NA,
   start.date = NA,
   end.date = NA,
   vector.ws = FALSE,
   fill = FALSE
)
Can't find the variable(s) date 
Error in checkPrep(mydata, vars, type = "default", remove.calm = FALSE,  : 


#converting to time series and using period.apply(), but can't find a way to group them by dive #, or use split() then convert to time series.
#create a time series data class from our data frame
> seal_dives$datetime<-as.POSIXct(seal_dives$datetime,tz="GMT")
> seal_xts <- xts(seal_dives, order.by=seal_dives[,1])
> seal_20<-period.apply(seal_xts$Depth, endpoints(seal_xts$datetime, "seconds", 20),  mean)

#split data by dive # but don't know how to do averages over 20 seconds
> seal_split<-split(seal_dives, seal_dives$diveNum)

Maybe there is a magical way to do this that I haven't found on the internet yet, or maybe I'm just doing something wrong in one of my methods.

Shawn68
  • 19
  • 4

1 Answers1

0

You can use floor_date function from lubridate to bin data every 20 seconds. Group them along with diveNum and D_phase to get average of other columns using across.

library(dplyr)
library(lubridate)

result <- df %>%
  group_by(diveNum, D_phase, datetime = floor_date(datetime, '20 sec')) %>%
  summarise(across(c(Depth, MSA, rate_s, HR), mean, na.rm = TRUE), .groups = 'drop')

result
Ronak Shah
  • 377,200
  • 20
  • 156
  • 213