Closed wcornwell closed 6 years ago
What do you mean by 'what aspect of the output'? As in which explanatory variables we are interested in?
Yeah exactly. I guess can save all the output, but it's a challenge to summarize.
Yeah. Okay.
I mean, IF the prelim results stay the same and brain residual does indeed remain important, I think that will be the first point-of-call.
In fact, the title could be 'brain size predicts urban birds' or something. It could be that important, to structure the paper around.
But, obviously everything would probably be good at some point (in supp files).
@coreytcallaghan remake::make()
and have a look!
hmm only problem is this coef estimates seem to be robust relative to phylo uncertainty, but not compared to the model complexity.
> phy_mod<-phylolm(response~ brain_residual ,data=analysis_data,phy=aus_bird_tree,
+ na.action = "na.fail", weights=(analysis_data$N/analysis_data$unique_localities))
>
> summary(phy_mod)
Call:
phylolm(formula = response ~ brain_residual, data = analysis_data,
phy = aus_bird_tree, na.action = "na.fail", weights = (analysis_data$N/analysis_data$unique_localities))
AIC logLik
2797 -1395
Raw residuals:
Min 1Q Median 3Q Max
-5.0109 -0.5418 0.6157 2.0784 6.0114
Mean tip height: 123.4092
Parameter estimate(s) using ML:
sigma2: 0.5068597
Coefficients:
Estimate StdErr t.value p.value
(Intercept) -2.2626 3.8369 -0.5897 0.5556
brain_residual -5.5214 1.3282 -4.1571 3.74e-05 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
bummer....
or it's all about brain size at a given "movement class"
phy_mod<-phylolm(response~
brain_residual + movement_class ,data=analysis_data,phy=aus_bird_tree,
na.action = "na.fail", weights=(analysis_data$N/analysis_data$unique_localities))
summary(phy_mod)
Call:
phylolm(formula = response ~ brain_residual + movement_class,
data = analysis_data, phy = aus_bird_tree, na.action = "na.fail",
weights = (analysis_data$N/analysis_data$unique_localities))
AIC logLik
2609 -1292
Raw residuals:
Min 1Q Median 3Q Max
-6.6043 -2.9241 -1.1711 0.5215 5.8026
Mean tip height: 123.4092
Parameter estimate(s) using ML:
sigma2: 0.3482471
Coefficients:
Estimate StdErr t.value p.value
(Intercept) -0.037105 3.211774 -0.0116 0.990787
brain_residual 5.071363 1.558464 3.2541 0.001209 **
movement_classdispersal and nomadic/irruptive -2.279564 0.213701 -10.6671 < 2.2e-16 ***
movement_classdispersal and partial_migrant -0.409836 0.208041 -1.9700 0.049353 *
movement_classdispersal and partial_migrant and nomadic/irruptive -0.138086 0.321416 -0.4296 0.667646
movement_classnomadic/irruptive -0.935808 0.301685 -3.1019 0.002023 **
movement_classnone -1.086247 1.193977 -0.9098 0.363350
movement_classpartial_migrant -0.648871 0.453567 -1.4306 0.153127
movement_classpartial_migrant and nomadic/irruptive -1.437336 1.500354 -0.9580 0.338494
movement_classtotal_migrant 0.233282 0.305251 0.7642 0.445065
movement_classtotal_migrant and nomadic/irruptive -0.940297 0.958845 -0.9807 0.327203
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
maybe real? what do you think?
this implies that brain residuals matters, but only within a given "movement class" and after accounting for phylogeny. What do you think? real or not?
done i think
Have to fit model across many trees. First step is to figure out what aspect of the output we're interested in.