Species 106: Pinus edulis

General Linear Models

Boosted Regression Trees

Generalized Linear Model Results

Data:

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

n(pres)                 =      619

n(abs)                  =      4387

n covariates considered =      9

total time for model fitting = 3.55min

Settings:

model family          : binomial

simplification method : AIC

Results:

number covariates in final model   : 25

Call:

glm(formula = response ~ I(bio05^2) + bio05 + bio08 + bio19 + 

    I(bio19^2) + di_all_model + I(bio09^2) + I(bio04^2) + bio09 + 

    I(di_all_model^2) + bio12 + I(bio12^2) + bio04 + I(bio08^2) + 

    bio05:bio08 + bio08:bio19 + bio05:bio19 + bio19:bio09 + bio19:bio12 + 

    bio12:bio04 + di_all_model:bio04 + bio05:bio04 + bio05:bio09 + 

    di_all_model:bio12 + bio08:bio12, family = model.family, 

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

Deviance Residuals: 

    Min       1Q   Median       3Q      Max  

-2.3448  -0.2144  -0.0473  -0.0033   5.7569  

Coefficients:

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

(Intercept)        -1.025e+02  2.613e+01  -3.921 8.81e-05 ***

I(bio05^2)         -8.347e-04  1.299e-04  -6.428 1.29e-10 ***

bio05               2.370e-01  8.111e-02   2.922 0.003483 ** 

bio08               1.623e-01  2.884e-02   5.629 1.82e-08 ***

bio19              -2.761e-01  5.285e-02  -5.225 1.74e-07 ***

I(bio19^2)         -6.879e-04  1.573e-04  -4.373 1.22e-05 ***

di_all_model        3.457e-01  8.016e-02   4.313 1.61e-05 ***

I(bio09^2)         -1.711e-04  3.024e-05  -5.656 1.55e-08 ***

I(bio04^2)         -3.461e-07  1.390e-07  -2.491 0.012753 *  

bio09              -8.545e-03  2.401e-02  -0.356 0.721928    

I(di_all_model^2)   7.219e-04  1.715e-04   4.209 2.56e-05 ***

bio12               1.861e-01  3.100e-02   6.005 1.92e-09 ***

I(bio12^2)         -1.191e-04  1.933e-05  -6.161 7.23e-10 ***

bio04               6.602e-03  3.544e-03   1.863 0.062469 .  

I(bio08^2)          6.417e-05  3.990e-05   1.608 0.107771    

bio05:bio08        -3.587e-04  1.042e-04  -3.442 0.000577 ***

bio08:bio19        -2.864e-04  1.054e-04  -2.717 0.006586 ** 

bio05:bio19         1.441e-03  1.797e-04   8.020 1.06e-15 ***

bio19:bio09        -2.080e-04  6.219e-05  -3.345 0.000822 ***

bio19:bio12         1.680e-04  7.348e-05   2.287 0.022209 *  

bio12:bio04        -1.144e-05  2.293e-06  -4.989 6.07e-07 ***

di_all_model:bio04 -4.614e-05  8.025e-06  -5.749 8.97e-09 ***

bio05:bio04         1.848e-05  6.003e-06   3.078 0.002085 ** 

bio05:bio09         1.284e-04  8.007e-05   1.603 0.108919    

di_all_model:bio12 -1.282e-04  7.087e-05  -1.809 0.070416 .  

bio08:bio12        -6.625e-05  3.812e-05  -1.738 0.082234 .  

---

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

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 3745.9  on 5005  degrees of freedom

Residual deviance: 1688.3  on 4980  degrees of freedom

AIC: 1740.3

Number of Fisher Scoring iterations: 9

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

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.7275207 

NULL Deviance                : 0.74828

Fit Deviance                 : 0.33725

Explained Deviance           : 0.41103

Percent Deviance Explained   : 54.93

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.18 

Confusion Matrix: 

         observed

predicted    1    0

        1  549  484

        0   70 3903

AUC                          :  0.9558 

Percent Correctly Classified :  88.93328 

Sensitivity                  :  0.8869144 

Specificity                  :  0.889674 

Kappa                        :  0.6033041 

True Skill Statistic         :  0.7765884 

   Calibration Statistics 

Intercept (general calibration)                            :  0.005475958 

Slope   (direction and variation in fit)                   :  1.008618 

Testa0b1 (overall reliability of predictors)               :  0.9775155 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9975337 

Testb1|a (refinement given correct calibration)            :  0.8311371 

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

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.714541  (sd 0.040364) 

NULL Deviance                : 0.74827 (sd 0.0023258)

Fit Deviance                 : 0.35769 (sd 0.041966)

Explained Deviance           : 0.39059 (sd 0.042432)

Percent Deviance Explained   : 52.195 (sd 5.629)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.168 (sd 0.010328) 

Confusion Matrix: 

         observed

predicted    1    0

        1  540  515

        0   79 3872

AUC                          :  0.94951  (sd 0.012039) 

Percent Correctly Classified :  88.13497  (sd 1.7938) 

Sensitivity                  :  0.8723427  (sd 0.031918) 

Specificity                  :  0.8826172  (sd 0.018459) 

Kappa                        :  0.5810917  (sd 0.048564) 

True Skill Statistic         :  0.7549599  (sd 0.042076) 

   Calibration Statistics 

Intercept (general calibration)                            :  -0.02306003  (sd 0.23292) 

Slope   (direction and variation in fit)                   :  0.9633308  (sd 0.20081) 

Testa0b1 (overall reliability of predictors)               :  0.4738153  (sd 0.3655) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.566982  (sd 0.26298) 

Testb1|a (refinement given correct calibration)            :  0.4405675  (sd 0.36817) 

Total time = 41.49 min

Boosted Regression Tree Modeling Results

Data:

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

n(pres)                 =      619

n(abs)                  =      4387

n covariates considered =      9

total time for model fitting = 0.83min

Settings:

random seed used             : 19680828

tree complexity              : 9

learning rate                : 0.0972

n(trees)                     : 300

model simplification         : cross-validation

n folds                      : 3

n covariates in final model  : 6

Relative influence of predictors in final model:

   Var   rel.inf

 bio08 27.817106

 bio19 21.662333

 bio05 21.257507

 bio09 11.014134

 bio12  9.879168

 bio15  8.369751

Important interactions in final model:

 v1 name1 v2 name2

  2 bio08  1 bio05

  6 bio19  1 bio05

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

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.9358965 

NULL Deviance                : 0.74828

Fit Deviance                 : 0.12119

Explained Deviance           : 0.62709

Percent Deviance Explained   : 83.804

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.27 

Confusion Matrix: 

         observed

predicted    1    0

        1  605   97

        0   14 4290

AUC                          :  0.9977 

Percent Correctly Classified :  97.78266 

Sensitivity                  :  0.9773829 

Specificity                  :  0.9778892 

Kappa                        :  0.903259 

True Skill Statistic         :  0.9552721 

   Calibration Statistics 

Intercept (general calibration)                            :  0.6293789 

Slope   (direction and variation in fit)                   :  2.271101 

Testa0b1 (overall reliability of predictors)               :  0 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9505222 

Testb1|a (refinement given correct calibration)            :  0 

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

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.7430845  (sd 0.043565) 

NULL Deviance                : 0.74827 (sd 0.0023258)

Fit Deviance                 : 0.33564 (sd 0.048734)

Explained Deviance           : 0.41263 (sd 0.049497)

Percent Deviance Explained   : 55.138 (sd 6.5586)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.28 (sd 0.027487) 

Confusion Matrix: 

         observed

predicted    1    0

        1  485  216

        0  134 4171

AUC                          :  0.95677  (sd 0.012222) 

Percent Correctly Classified :  93.00782  (sd 1.8017) 

Sensitivity                  :  0.7834479  (sd 0.041585) 

Specificity                  :  0.9507619  (sd 0.018441) 

Kappa                        :  0.6968933  (sd 0.065413) 

True Skill Statistic         :  0.7342098  (sd 0.049211) 

   Calibration Statistics 

Intercept (general calibration)                            :  0.02870463  (sd 0.26509) 

Slope   (direction and variation in fit)                   :  0.8586167  (sd 0.095931) 

Testa0b1 (overall reliability of predictors)               :  0.2144228  (sd 0.27527) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.3967291  (sd 0.32544) 

Testb1|a (refinement given correct calibration)            :  0.2660606  (sd 0.32604) 

Total time = 12.35 min

MARS

Random Forests

MARS Model Results

Data:

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

n(pres)                 =      619

n(abs)                  =      4387

n covariates considered =      9

total time for model fitting = 0.1min

Settings:

random seed used             : 123

mars degree                  : 1

mars penalty                 : 2

Summary of Model:

             nsubsets   gcv    rss

bio08              14 100.0  100.0

bio19              13  90.0   90.1

bio05              13  80.7   81.1

di_all_model        9  40.6   41.9

bio09               8  33.7   35.2

bio12               3  12.2   14.0

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

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.6924679 

NULL Deviance                : 0.74828

Fit Deviance                 : 0.38683

Explained Deviance           : 0.36144

Percent Deviance Explained   : 48.303

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.12 

Confusion Matrix: 

         observed

predicted    1    0

        1  532  658

        0   87 3729

AUC                          :  0.9328 

Percent Correctly Classified :  85.11786 

Sensitivity                  :  0.8594507 

Specificity                  :  0.8500114 

Kappa                        :  0.5081564 

True Skill Statistic         :  0.7094621 

   Calibration Statistics 

Intercept (general calibration)                            :  6.654607e-05 

Slope   (direction and variation in fit)                   :  1.000223 

Testa0b1 (overall reliability of predictors)               :  0.9999764 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9976369 

Testb1|a (refinement given correct calibration)            :  0.9950494 

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

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.6877934  (sd 0.040021) 

NULL Deviance                : 0.74827 (sd 0.0023258)

Fit Deviance                 : 0.40251 (sd 0.06353)

Explained Deviance           : 0.34576 (sd 0.064183)

Percent Deviance Explained   : 46.201 (sd 8.5295)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.129 (sd 0.0073786) 

Confusion Matrix: 

         observed

predicted    1    0

        1  530  624

        0   89 3763

AUC                          :  0.93205  (sd 0.01973) 

Percent Correctly Classified :  85.75745  (sd 1.0723) 

Sensitivity                  :  0.8561079  (sd 0.058077) 

Specificity                  :  0.8577636  (sd 0.011452) 

Kappa                        :  0.5205179  (sd 0.033819) 

True Skill Statistic         :  0.7138715  (sd 0.056213) 

   Calibration Statistics 

Intercept (general calibration)                            :  0.0212044  (sd 0.27573) 

Slope   (direction and variation in fit)                   :  1.017555  (sd 0.19699) 

Testa0b1 (overall reliability of predictors)               :  0.3834428  (sd 0.31737) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.5665344  (sd 0.23495) 

Testb1|a (refinement given correct calibration)            :  0.3417934  (sd 0.33897) 

Total time = 2.02 min

Random Forest Modeling Results

Data:

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

n(pres)                 =      619

n(abs)                  =      4387

n covariates considered =      9

total time for model fitting = 0.34min

Settings:

random seed used                       : 19680828

n covariates considered at each split  : 2

n trees                                : 1000

Relative performance of predictors in final model:

                   0       1 MeanDecreaseAccuracy MeanDecreaseGini

bio08        70.3952 62.2861              86.1833         128.4870

bio05        60.4599 56.1493              71.6299         101.0251

bio19        59.4401 58.4447              70.5978          90.7050

bio12        50.7442 43.2419              62.8779          70.4271

bio09        52.7913 44.1152              62.7271          82.2899

bio02        55.6196 26.5687              59.4317          52.4684

bio04        46.1199 40.6457              57.6993          60.2794

di_all_model 48.2337 19.3482              49.8899          44.8466

bio15        39.2548 38.7210              47.1583          53.8781

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

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.7737424 

NULL Deviance                : 0.74828

Fit Deviance                 : 0.29342

Explained Deviance           : 0.45485

Percent Deviance Explained   : 60.787

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.17 

Confusion Matrix: 

         observed

predicted    1    0

        1  561  389

        0   58 3998

AUC                          :  0.9684 

Percent Correctly Classified :  91.07072 

Sensitivity                  :  0.9063005 

Specificity                  :  0.9113289 

Kappa                        :  0.6649331 

True Skill Statistic         :  0.8176294 

   Calibration Statistics 

Intercept (general calibration)                            :  0.2340905 

Slope   (direction and variation in fit)                   :  1.217472 

Testa0b1 (overall reliability of predictors)               :  1.337379e-06 

Testa0|b1(incorrect calibration given correct refinement)  :  0.5155011 

Testb1|a (refinement given correct calibration)            :  2.468077e-07 

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

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.7684863  (sd 0.032403) 

NULL Deviance                : 0.74827 (sd 0.0023258)

Fit Deviance                 : 0.29761 (sd 0.028769)

Explained Deviance           : 0.45066 (sd 0.029641)

Percent Deviance Explained   : 60.223 (sd 3.8886)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.495 (sd 0.0091287) 

Confusion Matrix: 

         observed

predicted    1    0

        1  408  103

        0  211 4284

AUC                          :  0.96731  (sd 0.0097445) 

Percent Correctly Classified :  93.72743  (sd 0.75569) 

Sensitivity                  :  0.6590428  (sd 0.030089) 

Specificity                  :  0.9765241  (sd 0.0089234) 

Kappa                        :  0.6874868  (sd 0.032206) 

True Skill Statistic         :  0.6355669  (sd 0.028867) 

   Calibration Statistics 

Intercept (general calibration)                            :  0.2486845  (sd 0.31251) 

Slope   (direction and variation in fit)                   :  1.211483  (sd 0.2035) 

Testa0b1 (overall reliability of predictors)               :  0.3801963  (sd 0.36801) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.618126  (sd 0.27892) 

Testb1|a (refinement given correct calibration)            :  0.3344707  (sd 0.3958) 

Total time = 8.4 min