search

renikaul / YF_Brazil

Risk map of Yellow Fever in Brazil based on past cases
0 stars 1 forks source link

Figure to support "natural break in NHP richness" #42

Closed renikaul closed 6 years ago

renikaul commented 6 years ago

Need a figure with the 2 regions mapped and a histogram of NHP richness to support reasoning behind split. See John's comments Methods/Predictive Model/Two Separate Models paragraph.

The datas used for the national model above were further divided into two regional datasets, based on species richness of non-human primate reservoirs. We characterized municipalities as high reservoir richness (HRR, $>5$) or low reservoir richness (LRR, $\leq 5$). This division was determined by natural breaks in the distribution of the species richness data.

  • A HISTOGRAM AND/OR MAP SHOULD BE REFERENCED HERE TO SUPPORT THIS POINT.

To form contiguous regions, municipalities which were completely enveloped by a municipality belonging to another categorization I DON'T UNDERSTAND THIS were included in the same region as their surrounding neighbors. We estimated the performance of each model via the area under the receiver-operator curve (AUC) of the testing data.

  • INSERT SENTENCE ELABORATING ON INTERPRETATION OF AUC. ADD REFERENCE TO PAPER ON ROC ANALYSIS IN PATTERN RECOGNITION LETTERS -- I CAN LOOK UP FOR YOU IF NEED BE

Relative variable importance was assessed by the difference in median AUC of 100 permutations to the AUC of the original model, scaled to the largest decrease in AUC due to permutation of a single variable.

I think a histogram with an inset of the HRR/LRR poster map color coded would do the trick. Colors for the two regions should be consistent throughout figures. Thoughts?

mvevans89 commented 6 years ago

That works for me. I think a binary split would be more helpful than one that shows the number of species per municipality. Supplemental or in the main text?

renikaul commented 6 years ago

I don't think the histogram is accurate. We have some muni with >6 and <5 in the two regions (ie. surrounded by other area).

mvevans89 commented 6 years ago

15 municipalities that have more than 5 NHP species were included in the LRR region 28 municipalities that have less than 6 NHP species were included in the HRR region

Total number of municipalities per region: HRR: 210 LRR: 5351

Adding an additional figure in commit below in case we want something that better demonstrates this via a see through histogram. @renikaul Can you take a look and let me know your preference?

mvevans89 commented 6 years ago

New figure that shows how the municipalities were ranked and then combined based on geographic continuity