I am trying to design a competition for students focused on Bayesian understanding of ecology. Could I ask here for some pointers? I will have data on 2 sets of maybe physiologically linked parameters (from some research I plan to do next summer), and then offer the students to review qualitative descriptions of the link between them, like ‘Some plants having mycorrhiza have higher biomass than the average − 3 populations out of 5’ (see L. Carroll, The game of logic.) There will be other variables that might correlate with mycorrhiza more or less strongly than total biomass, and the students will have to make an educated guess as to which variable (s) have better predictive value (that is somewhat like what we start from in ‘real life’, if the subject has not been rigorously researched before).
For this task the participants will have a month. Then on some set day I will call those of them who have offered the more substantial explanations and give them exact probabilities of, say, (plant having mycorrhiza)/(plant having a higher than average biomass) etc. so they could compute the posterior probability of a GIVEN plant having mycorrhiza. Bonus points if they note that the variable(s) they have chosen work worse than some other one(s). For this task they will have several hours. What is perhaps more interesting to me is not the number they will give, but the way they will arrive at it.
I would gladly share the experimental data with anybody interested in a similar experiment. Is there some advice you can give me? Thank you.
I am trying to design a competition for students focused on Bayesian understanding of ecology. Could I ask here for some pointers? I will have data on 2 sets of maybe physiologically linked parameters (from some research I plan to do next summer), and then offer the students to review qualitative descriptions of the link between them, like ‘Some plants having mycorrhiza have higher biomass than the average − 3 populations out of 5’ (see L. Carroll, The game of logic.) There will be other variables that might correlate with mycorrhiza more or less strongly than total biomass, and the students will have to make an educated guess as to which variable (s) have better predictive value (that is somewhat like what we start from in ‘real life’, if the subject has not been rigorously researched before). For this task the participants will have a month. Then on some set day I will call those of them who have offered the more substantial explanations and give them exact probabilities of, say, (plant having mycorrhiza)/(plant having a higher than average biomass) etc. so they could compute the posterior probability of a GIVEN plant having mycorrhiza. Bonus points if they note that the variable(s) they have chosen work worse than some other one(s). For this task they will have several hours. What is perhaps more interesting to me is not the number they will give, but the way they will arrive at it. I would gladly share the experimental data with anybody interested in a similar experiment. Is there some advice you can give me? Thank you.