CODEWITHSUNDEEP
rrandomgpscategories
Danica
Danica

Reputation: 87

R: Random sampling a minimum number of observations from a category

I have location data from GPS collars and I am trying to simulate different scenarios based on the functionality of the collars in R. One of those simulations is that the collars tend to miss taking GPS points throughout the day (for various reasons). My data consists of 14 GPS points per day, and I want to randomly select (without replacement) a minimum number of 5 points, the possibility of a max 14.

In another simulation, I extracted 5 random points per day using this script from another thread here (R: Random sampling an even number of observations from a range of categories), but I do not fully understand all the different bits of the script that would allow me to alter it to get it to extract AT LEAST 5 points. Any advice most appreciated.

dat2 <- data.table(dat.r)
dat2.ss <- dat2[ , .SD[sample(1:.N,min(5,.N))], by=DayNo]

Output from data-frame (dat.r)

dput(head(dat.r, 20))
structure(list(Latitude = c(5.4118432, 5.4118815, 5.4115713, 
5.4111541, 5.4087853, 5.4083702, 5.4082527, 5.4078161, 5.4075528, 
5.407321, 5.4070598, 5.4064237, 5.4070621, 5.4070251, 5.4070555, 
5.4065127, 5.4065134, 5.4064872, 5.4056724, 5.4038751), Longitude = c(118.0225467, 
118.0222841, 118.0211875, 118.0208637, 118.0205413, 118.0206064, 
118.0204101, 118.0209272, 118.0213827, 118.0214189, 118.0217748, 
118.0223343, 118.0227079, 118.0226511, 118.0226916, 118.0220733, 
118.02218, 118.0221843, 118.0223316, 118.0198153), DayNo = c(1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L)), .Names = c("Latitude", "Longitude", "DayNo"), row.names = c(NA, 
20L), class = "data.frame")

Upvotes: 0

Views: 446

Answers (1)

jlhoward
jlhoward

Reputation: 59415

This should work:

library(data.table)
set.seed(1)    # for reproducible example
setDT(dat.r)[,.SD[sample(.N, sample(min(5,.N):min(.N,14),1))], by=DayNo]
#     DayNo Latitude Longitude
#  1:     1 5.411881  118.0223
#  2:     1 5.411154  118.0209
#  3:     1 5.407553  118.0214
#  4:     1 5.411843  118.0225
#  5:     1 5.411571  118.0212
#  6:     1 5.407062  118.0227
#  7:     1 5.408785  118.0205
#  8:     1 5.408370  118.0206
#  9:     2 5.406513  118.0221
# 10:     2 5.407025  118.0227
# 11:     2 5.406513  118.0222
# 12:     2 5.405672  118.0223
# 13:     2 5.403875  118.0198

This idea is that sample(x, n) takes a sample of size n from the vector 1:x (where x is a number, not a vector). So you want n to be itself sampled from 5:min(.N,14). I added the possibility that there are fewer than five points in a given day.

Upvotes: 2

Related Questions