Population likelihood interpretation? likelihood of total < likelihoods of both parts


#1

question: logically, how could the likelihood developed for a combined population (~0.05) be significantly different that for each half of the same population (~0.9 and ~0.75) ?

ref: graphics: https://github.com/cordphelps/brm ; first README figure, data mentioned for cluster one (red), first seasonal timeframe, week 23-25)

I have two (experimental and control) insect populations that yield ‘trapped insects’ over the course of time. I segment each of these populations into 3 clusters across 3 seasonal timeframes (9 separate sub-populations per transect)

I propose that the population drivers allow me to use the Oceanic Tool Complexity model described in McElreath (and https://bookdown.org/connect/#/apps/1850/access section 10.2) to assess the likelihood that the independent variable influences the number of trapped insects.

For the first seasonal population (weeks 23-25), I find that the likelihood that the dependent variable influences the number of trapped spiders in the third cluster (blue) is ~.75 and ~.05 for the first cluster (red).

I then partition this population by time of day (data was collected twice each day). I now find the half day (‘am’ and ‘pm’) likelihood for the first and third clusters is inverted compared to the analysis of the full day population. This seems weird to me. What could explain this?

‘brms’ package (version 2.6.0)
R version 3.3.3 (2017-03-06)
Platform: x86_64-apple-darwin13.4.0 (64-bit)
Running under: OS X El Capitan 10.11.6


#2

sorry : ‘dependent’


#3

I am sorry, I don’t think I can answer this question as I am lacking subject matter knowledge. It does not seem like a specific brms related issue, right? Does the same happen with out software?


#4

I was trying to pose the question such that you do not need subject matter knowledge, sorry for the unnecessary project detail! Here’s the elevator pitch:

Finding a high plausibility that an outcome variable is influenced by the combined effect of two independent variables, is there anything that I can assume about the same plausibility for two separate halves of the initial population?

Like “the mean plausibility of the two halves should equal the plausibility for the whole population” or something like that.

Thank you Paul!


#5

I don’t think such a statement can be made like that at least not without certain additional assumptions.