Format of model tables

[mcgregorian1]

Hi @teixeirak

[Edit]: updated tables with r2 values after changing REML=TRUE for best model

When presenting the best models for each scenario, how were you thinking would make the most sense?

1. Table with coefficients of each variable in each model in a list with r2

   • the table is not fully expanded here, in rmarkdown I can make it be such that everything is fit in the table by stretching it vertically

2. Table with just r2 and coefficients in a separate table (each scenario would have a separate coefficients table)

• this table is for 1999 (with REML=FALSE)

[mcgregorian1]

new table:

var_name	scenario1	scenario2	scenario3	scenario4
r2	0.13	0.24	0.21	0.25
(Intercept)
height.ln.m

etc.

[mcgregorian1]

@teixeirak here's an updated table, with REML=TRUE (since now these are just for the best model)

[teixeirak]

Very nice!

[teixeirak]

@mcgregorian1, as I review these issues I'm remembering that we had come up with the nice format above fro the best full models. Could we keep this format to present all best models? I'm not sure if it would fit as a single table for presentation, but I think that's the best solution to current table S3 (issue #47). We can split it up later if its too big.

We could also invert the table, if that would improve the size.

Let's add a row (column, if inverted) labeled dAIC.

[mcgregorian1]

@teixeirak I think this will take more thought. If I put everything together, then I get a table that's 37 columns across:

I could create a single table per top model like I did in previous comments, but this would ultimately mean there are 18 tables like this. I'm not sure if that'd be the best option or not, though.

[teixeirak]

I'm confused by this table. I'm envisioning one where the number of columns matches the number of top models. There would be one dAIC per model, presented the same way as r^2. Maybe this is something that's best to touch base about quickly in person today.

[mcgregorian1]

We want to have the model_vars along the top, and the models going down (so, inverting the table above when dAIC is added)

[mcgregorian1]

@teixeirak I think this is what you were envisioning?

[mcgregorian1]

I was thinking renaming the rows to be All years # 1 All years # 2 .. 1977 # 1 1977 # 2

what do you think?

[teixeirak]

Perfect!

[teixeirak]

Just a few small modifications:

• [x] please add a columns for 1964-66 and co-dominant trees populated with zeros in the relevant rows. This is to make it easier to interpret.
• [x] Add headings of year and canopy position above the relevant columns
• [x] move height ahead of canopy position

[teixeirak]

I actually think this table would be good in the main document. what do you think?

[teixeirak]

If formatting this table in R is a pain, we could copy and paste into an excel template.

[mcgregorian1]

I can put the headings in excel (I saved it as a csv but will just convert it to xlsx). It looks like it may be good for the main body but also I noticed NewPhyt says to keep all figures to an absolute minimum

[teixeirak]

I do see value in putting it in the main text, as it's a nice concise way of representing the full model results. Tables giving statistical model results tend to be a bit boring, but its really the best way to present this.

[teixeirak]

I've tentatively added it to the manuscript. I find it critical to refer to it a lot as I fill in results, so I think it should make the cut.

[mcgregorian1]

@teixeirak I believe we've made a decision about this, are we good to close this issue?

[teixeirak]

Yes.