Fix-up of Table S3

[teixeirak]

• [x] reverse the order of col 2-4,
• [x] rename col 2-4 MAT, MAT+MAP, MAT x MAP
• [x] rather than giving True-false, list p values (NA for MAT x MAP tests that weren't run because MAT+MAP wasn't sig.)

This covers Helene's comments in issue #82 .

[beckybanbury]

@teixeirak should the p.values presented be p values for MATMAP compared with MAT + MAP, MAT + MAP compared with MAT alone etc, or MATMAP, MAT+ MAP, MAT, each compared with the null model?

[teixeirak]

Let's go with the first option.

[beckybanbury]

@teixeirak I'm still having problems with this analysis unfortunately. Below is the table of p values:

Carbon flux	MAT	MAT + MAP	MAT x MAP	R-squared value
GPP	1.50E-13	8.20E-05	0.14	0.66
NPP	4.00E-11	0.45	0.018	0.48
ANPP	1.00E-19	0.035	0.46	0.45
ANPP stem	1.90E-10	0.83	0.021	0.26
ANPP foliage	1.10E-13	0.47	0.41	0.59
BNPP root	3.30E-06	0.96	0.056	0.29
BNPP fine root	0.0021	0.23	0.091	0.15
R auto	0.00016	0.041	0.34	0.71
R root	0.0011	0.11	0.11	0.25

I had coded this in reverse, starting with the full MAT x MAP model and working backwards, based on Valentine's idea of using the drop1 AIC function, but I think it may be wrong to use here. You can see from the table that where MAT x MAP is significant, MAT + MAP was not significant. If I code it the other way, i.e. start with MAT, and then add terms, testing for significance and dAIC at each stage, none of the interactions come out as significant, as they don't get past the additive model.

I'm pretty confused at this stage at what the correct approach is - can an interactive model only be significant if the additive model is also significant?

Sorry if I've messed this up again!

[teixeirak]

I think it's okay...right @ValentineHerr?

[beckybanbury]

Updated table is complete, providing this is statistically sound!

Carbon flux	MAT	MAT + MAP	MAT x MAP	R$^{2}$
GPP	<0.0001	<0.0001	NS	0.66
NPP	<0.0001	NS	0.018	0.48
ANPP	<0.0001	0.035	NS	0.45
ANPP stem	<0.0001	NS	0.021	0.26
ANPP foliage	<0.0001	NS	NS	0.59
BNPP root	<0.0001	NS	NS	0.29
BNPP fine root	0.0021	NS	NS	0.15
R auto	0.00016	0.041	NS	0.71
R root	0.0011	NS	NS	0.25

[teixeirak]

Looks great; thanks!