Species 092: Picea breweriana

General Linear Models

Boosted Regression Trees

Generalized Linear Model Results

Data:

C:\Withrow\Workspace\Model\glm_1\CovariateCorrelationOutputMDS_initial.csv

n(pres)                 =      46

n(abs)                  =      1410

n covariates considered =      5

total time for model fitting = 0.06min

Settings:

model family          : binomial

simplification method : AIC

Results:

number covariates in final model   : 6

Call:

glm(formula = response ~ bio06 + bio12 + I(bio06^2) + I(bio12^2) + 

    I(di_all_model^2) + I(bio04^2), family = model.family, data = dat, 

    weights = weight, na.action = "na.exclude")

Deviance Residuals: 

   Min      1Q  Median      3Q     Max  

-1.760  -0.007   0.000   0.000   3.511  

Coefficients:

                    Estimate Std. Error z value Pr(>|z|)    

(Intercept)       -3.006e+01  8.459e+00  -3.554  0.00038 ***

bio06             -5.095e-01  1.078e-01  -4.725 2.30e-06 ***

bio12              2.737e-02  1.244e-02   2.200  0.02780 *  

I(bio06^2)        -4.119e-03  9.906e-04  -4.158 3.21e-05 ***

I(bio12^2)        -9.986e-06  5.252e-06  -1.901  0.05727 .  

I(di_all_model^2) -2.197e-03  1.023e-03  -2.148  0.03169 *  

I(bio04^2)        -8.847e-08  5.607e-08  -1.578  0.11461    

---

Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 408.37  on 1455  degrees of freedom

Residual deviance: 149.16  on 1449  degrees of freedom

AIC: 163.16

Number of Fisher Scoring iterations: 12

============================================================ 

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.7103134 

NULL Deviance                : 0.28048

Fit Deviance                 : 0.10245

Explained Deviance           : 0.17803

Percent Deviance Explained   : 63.474

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.05 

Confusion Matrix: 

         observed

predicted    1    0

        1   44   85

        0    2 1325

AUC                          :  0.9844 

Percent Correctly Classified :  94.02473 

Sensitivity                  :  0.9565217 

Specificity                  :  0.9397163 

Kappa                        :  0.4785701 

True Skill Statistic         :  0.8962381 

   Calibration Statistics 

Intercept (general calibration)                            :  0.001465382 

Slope   (direction and variation in fit)                   :  1.002638 

Testa0b1 (overall reliability of predictors)               :  0.9998241 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9975949 

Testb1|a (refinement given correct calibration)            :  0.9852306 

============================================================ 

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.6541676  (sd 0.16957) 

NULL Deviance                : 0.28003 (sd 0.023422)

Fit Deviance                 : 0.12121 (sd 0.056769)

Explained Deviance           : 0.15882 (sd 0.049309)

Percent Deviance Explained   : 57.224 (sd 18.086)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.061 (sd 0.0073786) 

Confusion Matrix: 

         observed

predicted    1    0

        1   42   81

        0    4 1329

AUC                          :  0.9802  (sd 0.015302) 

Percent Correctly Classified :  94.1647  (sd 1.2104) 

Sensitivity                  :  0.92  (sd 0.13984) 

Specificity                  :  0.942547  (sd 0.011843) 

Kappa                        :  0.4758248  (sd 0.068003) 

True Skill Statistic         :  0.862547  (sd 0.13982) 

   Calibration Statistics 

Intercept (general calibration)                            :  -2.635555  (sd 8.3954) 

Slope   (direction and variation in fit)                   :  13.60862  (sd 38.962) 

Testa0b1 (overall reliability of predictors)               :  0.4123472  (sd 0.30076) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.662276  (sd 0.24844) 

Testb1|a (refinement given correct calibration)            :  0.3372489  (sd 0.31969) 

Total time = 0.5 min

Boosted Regression Tree Modeling Results

Data:

C:\Withrow\Workspace\Model\brt_1\CovariateCorrelationOutputMDS_initial.csv

n(pres)                 =      46

n(abs)                  =      1410

n covariates considered =      5

total time for model fitting = 0.15min

Settings:

random seed used             : 19680828

tree complexity              : 4

learning rate                : 0.0077

n(trees)                     : 600

model simplification         : cross-validation

n folds                      : 3

n covariates in final model  : 2

Relative influence of predictors in final model:

   Var  rel.inf

 bio06 53.89536

 bio12 46.10464

Important interactions in final model:

 v1 name1 v2 name2

  2 bio12  1 bio06

============================================================ 

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.809054 

NULL Deviance                : 0.28048

Fit Deviance                 : 0.073196

Explained Deviance           : 0.20728

Percent Deviance Explained   : 73.903

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.08 

Confusion Matrix: 

         observed

predicted    1    0

        1   44   56

        0    2 1354

AUC                          :  0.9938 

Percent Correctly Classified :  96.01648 

Sensitivity                  :  0.9565217 

Specificity                  :  0.9602837 

Kappa                        :  0.5847691 

True Skill Statistic         :  0.9168054 

   Calibration Statistics 

Intercept (general calibration)                            :  0.4632328 

Slope   (direction and variation in fit)                   :  1.55685 

Testa0b1 (overall reliability of predictors)               :  0.00998694 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9945472 

Testb1|a (refinement given correct calibration)            :  0.002403147 

============================================================ 

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.6412284  (sd 0.13713) 

NULL Deviance                : 0.28003 (sd 0.023422)

Fit Deviance                 : 0.12043 (sd 0.038243)

Explained Deviance           : 0.1596 (sd 0.034218)

Percent Deviance Explained   : 57.212 (sd 12.571)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.083 (sd 0.018738) 

Confusion Matrix: 

         observed

predicted    1    0

        1   42   59

        0    4 1351

AUC                          :  0.98198  (sd 0.0095773) 

Percent Correctly Classified :  95.67442  (sd 1.0182) 

Sensitivity                  :  0.92  (sd 0.16865) 

Specificity                  :  0.9581451  (sd 0.010854) 

Kappa                        :  0.5528657  (sd 0.079856) 

True Skill Statistic         :  0.8781451  (sd 0.16553) 

   Calibration Statistics 

Intercept (general calibration)                            :  0.1228843  (sd 0.60562) 

Slope   (direction and variation in fit)                   :  1.253544  (sd 0.41616) 

Testa0b1 (overall reliability of predictors)               :  0.6575434  (sd 0.25426) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.6619458  (sd 0.20972) 

Testb1|a (refinement given correct calibration)            :  0.5225266  (sd 0.26712) 

Total time = 1.5 min

MARS

Random Forests

MARS Model Results

Data:

C:\Withrow\Workspace\Model\mars_1\CovariateCorrelationOutputMDS_initial.csv

n(pres)                 =      46

n(abs)                  =      1410

n covariates considered =      5

total time for model fitting = 0.04min

Settings:

random seed used             : 123

mars degree                  : 1

mars penalty                 : 2

Summary of Model:

             nsubsets   gcv    rss

bio06              11 100.0  100.0

di_all_model        9  50.7   55.4

bio12               7  38.8   44.0

bio02               7  36.1   41.8

bio04               3  20.1   24.6

============================================================ 

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.6834515 

NULL Deviance                : 0.28048

Fit Deviance                 : 0.10832

Explained Deviance           : 0.17215

Percent Deviance Explained   : 61.378

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.08 

Confusion Matrix: 

         observed

predicted    1    0

        1   44   76

        0    2 1334

AUC                          :  0.9828 

Percent Correctly Classified :  94.64286 

Sensitivity                  :  0.9565217 

Specificity                  :  0.9460993 

Kappa                        :  0.5076304 

True Skill Statistic         :  0.902621 

   Calibration Statistics 

Intercept (general calibration)                            :  0.001704407 

Slope   (direction and variation in fit)                   :  1.002399 

Testa0b1 (overall reliability of predictors)               :  0.9998379 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9976302 

Testb1|a (refinement given correct calibration)            :  0.9858294 

============================================================ 

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.619751  (sd 0.19042) 

NULL Deviance                : 0.28003 (sd 0.023422)

Fit Deviance                 : 0.16829 (sd 0.177)

Explained Deviance           : 0.11174 (sd 0.16931)

Percent Deviance Explained   : 41.421 (sd 58.331)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.056 (sd 0.012649) 

Confusion Matrix: 

         observed

predicted    1    0

        1   40   93

        0    6 1317

AUC                          :  0.96244  (sd 0.054766) 

Percent Correctly Classified :  93.20534  (sd 2.4398) 

Sensitivity                  :  0.88  (sd 0.13984) 

Specificity                  :  0.9340365  (sd 0.022964) 

Kappa                        :  0.4341868  (sd 0.10465) 

True Skill Statistic         :  0.8140365  (sd 0.15052) 

   Calibration Statistics 

Intercept (general calibration)                            :  -1.609245  (sd 4.2196) 

Slope   (direction and variation in fit)                   :  22.07845  (sd 66.636) 

Testa0b1 (overall reliability of predictors)               :  0.4860174  (sd 0.32201) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.6825791  (sd 0.18503) 

Testb1|a (refinement given correct calibration)            :  0.3282137  (sd 0.32611) 

Total time = 0.24 min

Random Forest Modeling Results

Data:

C:\Withrow\Workspace\Model\rf_1\CovariateCorrelationOutputMDS_initial.csv

n(pres)                 =      46

n(abs)                  =      1410

n covariates considered =      5

total time for model fitting = 0.05min

Settings:

random seed used                       : 19680828

n covariates considered at each split  : 3

n trees                                : 1000

Relative performance of predictors in final model:

                    0        1 MeanDecreaseAccuracy MeanDecreaseGini

bio06          7.1729 104.1352              45.9991          22.0283

bio02         -9.0102  26.4508              15.5023           8.5699

bio12        -15.2046  37.4929              13.7845          10.2334

bio04          1.0123  21.5680              13.3507          11.2111

di_all_model -18.7026  32.6791              12.5529           4.3259

============================================================ 

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.6515255 

NULL Deviance                : 0.28048

Fit Deviance                 : 0.12324

Explained Deviance           : 0.15723

Percent Deviance Explained   : 56.06

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.03 

Confusion Matrix: 

         observed

predicted    1    0

        1   43   92

        0    3 1318

AUC                          :  0.9601 

Percent Correctly Classified :  93.47527 

Sensitivity                  :  0.9347826 

Specificity                  :  0.9347518 

Kappa                        :  0.4491789 

True Skill Statistic         :  0.8695344 

   Calibration Statistics 

Intercept (general calibration)                            :  -0.3673793 

Slope   (direction and variation in fit)                   :  0.9333836 

Testa0b1 (overall reliability of predictors)               :  0.2759684 

Testa0|b1(incorrect calibration given correct refinement)  :  0.1520537 

Testb1|a (refinement given correct calibration)            :  0.46939 

============================================================ 

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.6633857  (sd 0.14909) 

NULL Deviance                : 0.28003 (sd 0.023422)

Fit Deviance                 : 0.12342 (sd 0.056307)

Explained Deviance           : 0.15661 (sd 0.04784)

Percent Deviance Explained   : 56.491 (sd 17.982)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.473 (sd 0.016193) 

Confusion Matrix: 

         observed

predicted    1    0

        1   23   11

        0   23 1399

AUC                          :  0.96027  (sd 0.05314) 

Percent Correctly Classified :  97.66411  (sd 1.0819) 

Sensitivity                  :  0.51  (sd 0.21705) 

Specificity                  :  0.9921935  (sd 0.0078045) 

Kappa                        :  0.557391  (sd 0.20438) 

True Skill Statistic         :  0.5021935  (sd 0.21722) 

   Calibration Statistics 

Intercept (general calibration)                            :  0.1497536  (sd 0.8748) 

Slope   (direction and variation in fit)                   :  1.440079  (sd 0.92691) 

Testa0b1 (overall reliability of predictors)               :  0.5209303  (sd 0.26563) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.6206368  (sd 0.1945) 

Testb1|a (refinement given correct calibration)            :  0.4286169  (sd 0.30776) 

Total time = 0.38 min