I'm wondering if anyone might have some novel insights as to the best way to analyze the following data. It's a problem I've been thinking about in the back of my mind for a while, so I thought that I'd ask here. I have data that look like this:

     day event actor recipient
995    8   128     G         J
996    8   129     G         K
997    8   130     G         B
998    8   131     B         G
999    8   132     H         G
1000   8   133     G         H
1001   8   134     E         G
1002   8   135     G         J
1003   8   136     B         H
1004   8   137     G         H
1005   8   138     G         H
1006   8   139     B         J
1007   9     1     D         J
1008   9     2     A         J
1009   9     3     A         J
1010   9     4     H         J
1011   9     5     A         J
1012   9     6     D         H
1013   9     7     A         F
1014   9     8     D         J
1015   9     9     A         H
1016   9    10     D         J
1017   9    11     A         J
1018   9    12     F         J
1019   9    13     F         J
1020   9    14     F         H
1021   9    15     F         G
1022   9    16     F         H
1023   9    17     C         F
1024   9    18     C         G
1025   9    19     D         H

What you see here is an extract of a R dataframe. The first column being the rownumber of the df, then the four variables. These data start at day1 and end at day 22. There are between 13 and 215 'events' on each day - each event is a separate behavioral event. Higher number events occur later in time than earlier numbered events. Individuals are in the 'actor' and 'recipient' variables. The data are available in csv format here:

There are 11 individuals (A - K). One thing you'll notice is that recipients tend to be lower down the alphabet, and actors tend to be higher up the alphabet.

A key question I'm interested in working out a methodology to address is to see if the likelihood of becoming an actor increases if an individual has recently been a recipient. You can see this on line 997 that G-B and then B-G occur followed by H-G and G-H. An individual recipient doesn't have to appear on the next line to count as having an increased likelihood of appearing as an actor - I'm interested in the decay in the probability of this occurring over events (but not continuing to the next day).

Further, I don't think this will be true for all individuals, so I am keen to test which individuals it is true for.

Lastly, I am interested to know if an individual who has recently been a recipient but is now an actor is paired with more often an individual of a higher or lower letter than themselves.

I hope that these questions are clear and make sense. I obviously don't expect a full analysis. But I would be keen to hear of ideas for this type of analysis. I believe that examining some Markov processes may be useful but I am interested in hearing about other ideas.

  • $\begingroup$ you need to start by creating some basic stats on individuals do/don't get converted from recipients to actors. I think you'll have a hard time getting people here to do that for you. I would suggest tackling the basic stats and see how far you get...if you run into problems then post your code and intent-it'll give folks here something to work with and it will demonstrate that you giving your problem some effort. $\endgroup$ May 19 '15 at 17:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.