So far we have used an example that came pre-formatted into a capture history dataframe. Usually however our field data are going to be in a less structured format, typically a series of days or other encounter occasions and a list of animals (by id number or other label) that are captured each occasion. So this typical structure is something like
Occasion Id
1 7
1 18
1 22
....
2 6
2 18
2 3
2 22
...
etc
What we want instead is a matrix of capture histories, where the rows are the indiviual animal (the id labels may or may not be needed), and the columns are entries 1 indicating the animal was captured on that occasion, or 0, was not. For instance for 5 occasions something like
10100
00100
10000
etc
By definition of course we are never going to observe
00000
which are animals that were never captured.
There are a number of ways of accomplishing this including
Constructing the matrix by hand (only recommended for very small problems)
Constructing the matrix using Excel or Access pivot tables (not my department)
R or other program to do the job (approach here)
I have written some code in R that reads data from a csv file containing capture dates and identification labels for 30 animals captured at least once over 10 days, and converts it to a capture history matrix where "occasion" can be defined by the user (e.g, daily, 2-day periods, etc.). The data are located here and the program is here. The script reads the raw csv file, converts dates to an R date-time object, and runs the data frame through a user-defined function cap.convert() to create capture histories given a specified capture occasion length in days (in the examples, 1- or 2- day intervals). The data are then read by the RMark function mark and run through a series of closed models (for this example, the estimates of N are extremely close to 30, suggesting that all or nearly all of the population was captured at least once.