Papers with Multiple Fire Severities

[jwillms]

Here is a possible idea for getting more data.

Some papers reported data for multiple fire severities. As in, in a single paper, there would be data for "low severity" fires, "mild severity" fires, and "high severity" fires.

In these cases, I pulled "mild severity" only. What if we pulled data for all three?

I think this may be a good idea. This is because fires from each study across the entire analysis were not uniform in severity--If there was a fire, I used the data (I did not exclude severe fires from the analysis, nor did I exclude low intensity fires). I only did so in individual papers when there was a "mild" option.

I think the way I did it before is now a mistake. We should pull the data for every fire severity, even if they are reported in a single paper. This would effectively increase the number of studies for 3-5 papers (I don't remember the exact number of papers that were like this).

In short, if we do this, we get more data from the papers we already have.

Thoughts?

[jwillms]

I am adding data for a new paper, Fornwalt, Kaufmann, and Stohlgren (which I will add to the "new" papers). This paper illustrates the point I made in the previous comment--it has multiple sites with multiple fire severities. I am going to pull data from every site for this paper, as per my recommendation in the previous post. If this is a bad idea, we can just throw out the "severe" and "low severity" data.

[jwillms]

The problem with this idea is that the severity of the fires absolutely has an effect on the change in richness and cover. Fornwalt, Kaufmann, & Stohlgren (one of the new papers) states this explicitly on page 2688. However, as I stated before, we are already comparing fires of different severities across the entire study, so I think we should do it.

[dschwilk]

If these report separate experiments, then yes, we can include them separately. You will need to create an id that distinguishes the experiments reported in a single publication.

[dschwilk]

Are you including some index of fire severity as a covariate? We have not done any covariate anaysis yet but it seems to me that year since treatment, fire severity, and vegetation type (tricky) are likely options. Make sure these are recorded with each paper.

[jwillms]

They generally do report separate experiments. I will need to go through each paper again to see which ones this applies to. There is some good potential here for getting a bit more data. Working on this.

[jwillms]

For the fire severity covatiate, should I simply report a (somewhat arbitrary) "low" "medium" or "high" severity for each fire? This would be the simplest way.

[dschwilk]

That is ok, but you should have some guideline to explain the method. As long as there are criteria.

[dschwilk]

@jwillms will send me specific list of questions and I will go through pubs as needed to answer