Refine statistical model

forc-db / Global_Productivity

Creative Commons Attribution 4.0 International

2 stars 0 forks source link

Refine statistical model #5

Closed teixeirak closed 5 years ago

teixeirak commented 5 years ago

@beckybanbury,

The current model (discussed in issue #2, #4, coded here ) is:

dependent variables:

all ForC flux variables with >30(?) records (run independently)

fixed effects:

[climate]
- currently we’re looking at latitude, MAT, MAP (run independently)
dominant.veg
- leaf type : broadleaf/ needleleaf,
- leaf phenology : evergreen/ deciduous

random effects:

plot within area

constraints:

stand age >=100 or "999"
min.dbh <=10cm
remove disturbed/ managed (?- I think)

First step is to identify the most important refinements needed.

beckybanbury commented 5 years ago

@teixeirak I'm having a look through the variables used; on the code + issues where this is discussed it looks like some of the flux variables have been combined (e.g. ANPP_0, ANPP_1 and ANPP_2 combined to ANPP) - is this the current approach, or is the aim to run them all independently? I can see there are advantages and disadvantages to both approaches!

teixeirak commented 5 years ago

As mentioned in person, influential methodological differences could serve as covariates in our analysis. Less influential methodological differences can be merged. Data collected using sub-standard methods should be dropped.

Note that ANPP_0 (or others with the _0 subscript) indicates that we don't know whether measurements meet standards of ANPP_1 or ANPP_2 -- in many cases because we haven't looked it up. Thus, a lot of the ANPP_0 records could be classified.

teixeirak commented 5 years ago

I think this issue is out of date.