After selecting the climate variables, they were applied in a logistic model, and the model’s performance metrics were evaluated.

library(gsheet)
final_df = gsheet2tbl("https://docs.google.com/spreadsheets/d/1BymEWPRq0ERQ3umD5IOvtK-wa4NMoq4C/edit?usp=sharing&ouid=116573839171815179218&rtpof=true&sd=true")

Logistic model

Example of a model: consider changing the threshold and other climate variables.

library(rms)
library(rmda)
library(pROC)
library(caret)

final_df$e_a

NULL

library(rms)
dd <- datadist(final_df)
options(datadist = "dd")

m_logistic <- lrm(epidemic ~PRECTOTCORR_14_23 + e_a_10_19,data = final_df, x=TRUE, y=TRUE)
summary(m_logistic)

             Effects              Response : epidemic 

 Factor            Low    High   Diff.   Effect    S.E.    Lower 0.95
 PRECTOTCORR_14_23 2.5535 6.6008 4.04720 -0.024322 0.19108 -0.39883  
  Odds Ratio       2.5535 6.6008 4.04720  0.975970      NA  0.67110  
 e_a_10_19         1.8623 2.2711 0.40876 -0.523650 0.26426 -1.04160  
  Odds Ratio       1.8623 2.2711 0.40876  0.592360      NA  0.35289  
 Upper 0.95
  0.3501900
  1.4193000
 -0.0056997
  0.9943200

plot(Predict(m_logistic),conf.int = TRUE)

pred <- predict(m_logistic, final_df, type = "fitted")
final_df$epidemic <- as.numeric(as.character(final_df$epidemic))
compare <- data.frame(1, final_df$epidemic, pred)
library(PresenceAbsence)
optimal.thresholds(compare)

         Method      pred
1       Default 0.5000000
2     Sens=Spec 0.5200000
3  MaxSens+Spec 0.5800000
4      MaxKappa 0.5800000
5        MaxPCC 0.5800000
6  PredPrev=Obs 0.5100000
7       ObsPrev 0.5309278
8      MeanProb 0.5309278
9    MinROCdist 0.5200000
10      ReqSens 0.4500000
11      ReqSpec 0.6000000
12         Cost 0.5800000

library(caret)
confusionMatrix(data = as.factor(as.numeric(pred > 0.76)),  mode= "everything",  reference = as.factor(final_df$epidemic))

Confusion Matrix and Statistics

          Reference
Prediction   0   1
         0  90 103
         1   1   0
                                          
               Accuracy : 0.4639          
                 95% CI : (0.3922, 0.5368)
    No Information Rate : 0.5309          
    P-Value [Acc > NIR] : 0.9738          
                                          
                  Kappa : -0.0103         
                                          
 Mcnemar's Test P-Value : <2e-16          
                                          
            Sensitivity : 0.9890          
            Specificity : 0.0000          
         Pos Pred Value : 0.4663          
         Neg Pred Value : 0.0000          
              Precision : 0.4663          
                 Recall : 0.9890          
                     F1 : 0.6338          
             Prevalence : 0.4691          
         Detection Rate : 0.4639          
   Detection Prevalence : 0.9948          
      Balanced Accuracy : 0.4945          
                                          
       'Positive' Class : 0               
                                          

library(pROC)