Subset a data frame by randomly selecting values based on two columns

Question

I have a data frame that I want to subset by randomly selecting 25 values of ID based on spp == cat and 25 values of ID based on spp == dog .

Here is my example data:

ID  spp category    prop
1   cat small_mam   0.99
2   cat small_mam   0.8
2   cat birds       0.15
3   dog large_mam   1
4   dog med_mam     0.4
4   dog emu         0.6
10  dog med_mam     0.8
10  dog birds       0.2
12  dog reptiles    1
13  dog large_mam   1
14  dog large_mam   1
15  dog large_mam   1
27  cat birds       0.2
28  cat small_mam   1
29  cat small_mam   0.75
29  cat birds       0.25
30  cat small_mam   0.7
30  cat birds       0.2

ID values for spp are unique meaning that cat and dog never have the same ID value. ID ranges from 1 to 696 but is not necessarily unique, this is because ID can be composed of up to 7 categories so randomly sub-setting 25 rows for each species does not work.

The context behind this question is that I will be drawing 1000 random samples of 25 cat and 25 dog scats (UID = the scat ID number) for a bootstrap calculation of dietary overlap using the piankabio function in package(pgirmess).

Thanks in advance for any help.

I am using R version 3.1.3

Answer 1

With data.table you could do it as follows:

library(data.table)
subdf <- setDT(mydf)[, sample(ID, 5), by = spp]

On the example data you provided this gives:

 > subdf spp V1 1: cat 27 2: cat 30 3: cat 2 4: cat 28 5: cat 30 6: dog 10 7: dog 14 8: dog 12 9: dog 4 10: dog 15 

When you want to keep all columns (which I suppose you want to), you can do:

subdf <- setDT(mydf)[, .SD[sample(.N, 5)], by = spp]

which gives:

 > subdf spp ID category prop 1: cat 29 small_mam 0.75 2: cat 1 small_mam 0.99 3: cat 2 birds 0.15 4: cat 30 small_mam 0.70 5: cat 28 small_mam 1.00 6: dog 14 large_mam 1.00 7: dog 15 large_mam 1.00 8: dog 13 large_mam 1.00 9: dog 10 birds 0.20 10: dog 4 med_mam 0.40 

---

Note: I used a sample of 5 for explanatory reasons as the example dataset is not large enough to draw a sample of 25.

---

In respons to your comment, you can achieve that with:

setDT(mydf)
set.seed(4321)
newdf <- mydf[mydf[, .(ID = sample(unique(ID), 5)), by = spp], on = c("spp", "ID")]

which gives:

 > newdf ID spp category prop 1: 27 cat birds 0.20 2: 29 cat small_mam 0.75 3: 29 cat birds 0.25 4: 2 cat small_mam 0.80 5: 2 cat birds 0.15 6: 1 cat small_mam 0.99 7: 28 cat small_mam 1.00 8: 14 dog large_mam 1.00 9: 13 dog large_mam 1.00 10: 15 dog large_mam 1.00 11: 4 dog med_mam 0.40 12: 4 dog emu 0.60 13: 12 dog reptiles 1.00 

Explanation : With mydf[, .(ID = sample(unique(ID), 5)), by = spp] you create an index data.table with 5 unique ID's for each category of spp . Then you do a join on spp & ID where you use this index-data.table to select the part of mydf with these ID's.