Timeline for What statistics can be used to analyze and understand measured outcomes of choices in binary trees?
Current License: CC BY-SA 4.0
18 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| S Apr 28, 2021 at 12:10 | history | bounty ended | CommunityBot | ||
| S Apr 28, 2021 at 12:10 | history | notice removed | CommunityBot | ||
| Apr 26, 2021 at 16:01 | answer | added | Henrik | timeline score: 6 | |
| Apr 26, 2021 at 15:06 | comment | added | Jake Westfall | This could possibly be a good fit for a class of statistical models studied in psychology called multinomial processing trees. @henrik knows about these model, maybe he could provide his expert opinion on whether that would make sense here | |
| Apr 26, 2021 at 9:14 | comment | added | S Pr | Thanks guys. You've been a great help. Enough to answer the question; if you'd like to post an answer for posterity and archiving for future searches, I'll gladly accept it! | |
| Apr 24, 2021 at 3:13 | comment | added | Dan Slone | I agree @jbuddy_13. It is literally a hierarchical system! | |
| Apr 23, 2021 at 19:52 | comment | added | jbuddy_13 | Whatever you end up doing, I think a hierarchical Bayesian approach would scale best as you add more layers in your tree. See Statistical Rethinking youtube.com/playlist?list=PLDcUM9US4XdNM4Edgs7weiyIguLSToZRI. | |
| Apr 23, 2021 at 9:41 | comment | added | S Pr | Hi @DanSlone! We expect subjects to reach the lowest level. If not, we could still record those that don't reach the end within a time-limit. We also record this all on camera and can determine the time in seconds required to reach the nodes, or 'chambers' as you say. The arena could be split into discrete zones, or the subject's progress from start to finish could be tracked (continuously, from 0 to 1). There is room for interpretation and design of the experiment, so I would like the analysis and statistics to be worked out beforehand. I foresee it being very interesting and contentious! | |
| Apr 23, 2021 at 3:15 | comment | added | Dan Slone | Still thinking about this @S Pr .... are you expecting that the subject will end up in one of the lowest levels every time, or could they stay in any of the chambers? That would greatly influence the analysis. Assuming there will be a time limit for the final determination. Alternatively you could also mark the time they take to travel from node to node, but that would be much more complicated. | |
| Apr 21, 2021 at 10:36 | comment | added | S Pr | First: thank you so very much! Great comments. Conditional probabilities and Bayesian network approach seems to fit, will look into it further too. I am heavily interested in interaction effects as @Dan Slone mentions. Can these be accounted for with conditional probabilities or Bayesian network analysis? As for limitations... no, I am NOT using the same animal from run-to-run (they are independent tests) and there are no major sample size limitations in my opinion. Say n=30 animals for each arena, so every tree (node permutation) would consist of 30 trials using single animals. More possible! | |
| Apr 20, 2021 at 15:02 | history | tweeted | twitter.com/StackStats/status/1384522817753034754 | ||
| Apr 20, 2021 at 14:12 | comment | added | Mikhail Popov | I think @DemetriPananos is onto something. Conditional probabilities might get you far here and you could either calculate them by hand, or describe the experiment as a Bayesian Network and estimate them using pomegranate (Python, free), bnlearn (R, free), or AgentaRisk (GUI, paid). | |
| Apr 20, 2021 at 14:03 | comment | added | Dan Slone | I think you must be interested in synergistic effects (e.g. the effect of A&B is greater than A + B) or antagonism (A&B has less attraction than A or B alone). So, you would be looking at interaction effects. Otherwise, pairing each grouping in a simple olfactometer would be the straightforward way to go. | |
| Apr 20, 2021 at 13:54 | comment | added | Demetri Pananos | Are there any limitations on your sample size? Naively, with enough data you could just look at conditional probabilities e.g. probability the animal choose A given A and B are mixed at level 2. Additionally, will you be using the same animals from run to run? | |
| S Apr 20, 2021 at 11:04 | history | bounty started | S Pr | ||
| S Apr 20, 2021 at 11:04 | history | notice added | S Pr | Draw attention | |
| Apr 17, 2021 at 13:43 | review | First posts | |||
| Apr 17, 2021 at 21:26 | |||||
| Apr 17, 2021 at 12:57 | history | asked | S Pr | CC BY-SA 4.0 |