search

forc-db / Global_Productivity

Creative Commons Attribution 4.0 International
2 stars 0 forks source link

Ratio regressions #106

Closed beckybanbury closed 3 years ago

beckybanbury commented 3 years ago

@teixeirak I realised I made a silly error and I ran the ratio regressions including an altitude term, even though we decided a while ago to exclude that from the models. In addition, I actually uploaded the wrong figure into the SI, from a script before I made the changes Helene suggested (calculating variables where we have the two other variables). I've now corrected these mistakes - the graph is here. As you can see, several ANPP:NPP relationships are now significant, but not BNPP:NPP or ANPP:BNPP, despite the fact that these all now use the same dataset (the slight differences in the number of samples are because I removed outliers, but the significance doesn't change if I leave outliers in - would also be good to get your thoughts on whether we should remove them or not). This is not what I expected! What do you think?

all

beckybanbury commented 3 years ago

NB ANPP:BNPP and foliage:stem are log transformed values because I haven't yet figured out how to back transform and keep the line fitting/transform the y axis

teixeirak commented 3 years ago

Thanks, and no problem-- I hadn't updated any results based on that plot (or any results in the manuscript, for that matter). That's one thing you could work on now; I'm going to have to stop for now, and probably for the rest of the day.

beckybanbury commented 3 years ago

What do you think re removing outliers? This is the plot where outliers are not removed:

all

I'm tempted to not remove, as I'm never sure about justification for removing outliers.

I also still can't work out how to transform the log values/yaxis for those plots of log ratios - is it okay as it is where I have just noted log(___) in the subheadings?

beckybanbury commented 3 years ago

Sorry Krista - ignore all this, I've just confused myself! I've just gone back through previous issues and I think this is all fine, and I resolved it fully previously. I will update the graph and results now