Closed teixeirak closed 4 years ago
Table 4- individual predictor tests
- [x] @mcgregorian1 , why is dAIC the same for DBH and H? Is this correct?
I can confirm that the dAIC for DBH and H is coming directly from the code, and after testing for dAIC all years individually (running each line of code), I can also confirm that the results are like what we're seeing. My only guess is that because DBH and height are linked intrinsically, we're getting the same result? But I agree it's odd we're seeing the exact same numbers for everything, including likelihood and the other stats that AIC yields.
To test it (in case Valentine wants to take a look at it):
trees_all_sub <- read.csv("manuscript/tables_figures/trees_all_sub.csv", stringsAsFactors = FALSE)
#one line is height, the other is DBH
models <- c("resist.value ~ year+(1|sp/tree)", "resist.value ~ year+(1|sp/tree)+height.ln.m")
# models <- c("resist.value ~ year+(1|sp/tree)", "resist.value ~ year+(1|sp/tree)+dbh.ln.cm")
lmm_all <- lapply(models, function(x){
fit1 <- lmer(x, data = model_df[[j]], REML=FALSE,
control = lmerControl(optimizer ="Nelder_Mead"))
return(fit1)
})
names(lmm_all) <- models
var_aic <- aictab(lmm_all, second.ord=TRUE, sort=FALSE) #rank based on AICc
#the individual year models are all subsets of the full, so it would make sense that if we're seeing the identical results for the full, we'd see it for the individual
- [x] @mcgregorian1 , did we use Y in the null models after it came out non-significant? (I think either way is justifiable, but make sure table is correct)
I'm not sure what Y is referring to here?
Y refers to year.
Ah right. Yes we did keep Y in for the null models when doing all years together because we figured that was the point for testing across years
Regarding height and DBH AICs, it wasn’t this way in previous iterations. While H is derived from DBH, they’re not so closely linked in that the allometey is different for every species. Perhaps confirm nothing went wrong in that step in a more recent iteration? Could one equation have been used for all species?
I'm a little confused by it because I double-checked the data and the H:DBH ratio is not consistent throughout, which is what I'd expect. I'll look at the equations that made DBH and height
Ok, the dAIC change happened on 19 Oct with my "new results" commit. I think that was where I had realized the height equations weren't correct so I changed them to be what they should. I'm going to check those equations now.
@teixeirak I realized some of the heights from NEON weren't being included in the regression analyses. But even after changing that and fixing the regression equations, I'm still getting the same dAIC, albeit a slightly different coefficient than before. At this point I'm not sure of the reason.
I'm in the process of changing the full analysis script so it runs and reads better (and is easier for someone else to follow). I should have that finished tonight.
I'm a little confused by it because I double-checked the data and the H:DBH ratio is not consistent throughout, which is what I'd expect. I'll look at the equations that made DBH and height
We would not expect it to be constant, just given the nature of the height allometry equations. So don't worry about that.
@teixeirak I realized some of the heights from NEON weren't being included in the regression analyses. But even after changing that and fixing the regression equations, I'm still getting the same dAIC, albeit a slightly different coefficient than before. At this point I'm not sure of the reason.
I'm in the process of changing the full analysis script so it runs and reads better (and is easier for someone else to follow). I should have that finished tonight.
Does this change the full model results?
I can confirm the full model results do not change because the candidate traits themselves do not change.
Right, but if some of the predicted heights are different, that changes model coefficients and perhaps variables included, right?
Yes I'm going through the rest of the code now
The updated coefficients are now on github (they didn't change that much, mostly by a few hundredths). I'm finishing clarifying the code and will upload soon
The updated coefficients are now on github (they didn't change that much, mostly by a few hundredths). I'm finishing clarifying the code and will upload soon
Okay. So there were no changes to the other coefficients, etc?
Could you please update tables 4 and 5?
My bad for not being specific. I've now updated tables 4 and 5. The coefficients that changed in table4 were only the ones for height and twi*height, whereas for table 5 the only changes were slight coefficient and Intercept changes for the models that had height as a variable.
@mcgregorian1, I believe I've done all "my part" for now, and that everything left for you is referenced with check boxes above. A lot of it is fairly minor clean-up, with probably the most substantive change being changes to Fig. 2.
Once the substantive changes above are done, we can send to coauthors for final review and get ready to submit!
Ok sounds good. I might be able to get to figure 2 in the next couple days otherwise it should be good by this weekend.
@teixeirak Figure 2 has now been fully updated. I will get to the other parts of this issue soon, but I think it is good to send to coauthors again unless you think something else needs to be done beforehand.
Looks great; thanks! I'm going to take a pass now and try to send to coauthors today.
Table 3-
- [ ] formatting
For formatting, I put in the appropriate variable symbolization but there's something going on with the commands to make it bold and differently-aligned, I think, that's currently causing it to not render like we want.
For formatting, I put in the appropriate variable symbolization but there's something going on with the commands to make it bold and differently-aligned, I think, that's currently causing it to not render like we want.
this is fixed.
remaining items moved to #82.
@mcgregorian1, our checklist in issue #48 was getting intractable/ out of date. Here's a new list:
Throughout
Figures & Tables
General:
FIg. 1-
Fig. height-
Table 1-
Table 2-
Table 3-
Table 4- individual predictor tests
Table 5- best full models
manuscript
General-
Abstract-
[x] work on abstract (mainly @teixeirak )
[ ] add supplementary info tables list
supplementary info
table S3-
table S4 -
table S5-
Figure S3-