**phenolology 🐢** @susannah · Mar 14, 2018, 17:26

**phenolology 🐢** @susannah · Mar 14, 2018, 17:26

phenolology 🐢 @susannah

Mar 14, 2018, 17:26

sciencing! phenology + treesex Show more

I'm building a model to figure out how much heat sum lodgepole pine trees need to shed pollen from their male cones and have receptive female cones. I have data from trees sourced from all over BC grown together at a bunch of different sites also in BC. Some of the trees are clones of each other and I have a bunch of years of data. So I'm making a hierarchical bayesian model.

**phenolology 🐢** @susannah · Mar 14, 2018, 17:30

**phenolology 🐢** @susannah · Mar 14, 2018, 17:30

Mar 14, 2018, 17:30

phenolology 🐢 @susannah

sciencing! phenology + treesex Show more

To help me understand the model and fix bugs in my code, I'm building the model one layer at a time and running it on simulated data. Yesterday I added multiple years and individual effects to the simulated data so I could add individual effects to the model. I ran a mixed model with lme4 in R just to try and it works better than a model without individual and year effects - most especially it's better at estimating the steepness of the curve.

**phenolology 🐢** @susannah · 2018-03-14T17:35:35Z

phenolology 🐢 @susannah

sciencing! phenology + treesex Show more

The green lines are the "real" model that I simulated data from, with individual effects. The red line is the new mixed model, but I haven't figured out how to extract the individual effects from lme4 and add them to the graph curves yet. And the black line is what happens when you have individual effects, but only one year.

Today I'm going to figure out how to get individual effects out of glmer (oops, lme4 above was a mistake) models and graph them.

Mar 14, 2018, 17:35 · Web · 0 · 1

**phenolology 🐢** @susannah · Mar 14, 2018, 17:38

**phenolology 🐢** @susannah · Mar 14, 2018, 17:38

Mar 14, 2018, 17:38

phenolology 🐢 @susannah

sciencing! phenology + treesex Show more

I also want to try updating my hierarchical bayesian model that I'm doing with stan via the rethinking package and running the simulated data through it.