Species 018: Abies lasiocarpa var. arizonica

General Linear Models

Boosted Regression Trees

Generalized Linear Model Results

Data:

G:\Results01\Withrow\Model\glm_1\CovariateCorrelationOutputMDS_initial.csv

n(pres)                 =      204

n(abs)                  =      4797

n covariates considered =      4

total time for model fitting = 0.22min

Settings:

model family          : binomial

simplification method : AIC

Results:

number covariates in final model   : 7

Call:

glm(formula = response ~ bio16 + bio02 + bio10 + I(bio02^2) + 

    I(bio10^2) + bio02:bio10 + bio16:bio10, family = model.family, 

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

Deviance Residuals: 

    Min       1Q   Median       3Q      Max  

-1.9795  -0.0895  -0.0095  -0.0008   3.5753  

Coefficients:

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

(Intercept) -3.773e+01  9.008e+00  -4.188 2.81e-05 ***

bio16       -1.063e-02  1.118e-02  -0.951 0.341511    

bio02        5.820e-01  1.181e-01   4.928 8.29e-07 ***

bio10       -3.100e-01  5.763e-02  -5.380 7.46e-08 ***

I(bio02^2)  -1.993e-03  4.326e-04  -4.606 4.11e-06 ***

I(bio10^2)  -1.269e-04  9.270e-05  -1.369 0.170950    

bio02:bio10  1.287e-03  3.390e-04   3.798 0.000146 ***

bio16:bio10  4.027e-04  8.599e-05   4.683 2.83e-06 ***

---

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

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 1704.9  on 5000  degrees of freedom

Residual deviance:  769.5  on 4993  degrees of freedom

AIC: 785.5

Number of Fisher Scoring iterations: 11

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

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.6335189 

NULL Deviance                : 0.34091

Fit Deviance                 : 0.15387

Explained Deviance           : 0.18704

Percent Deviance Explained   : 54.864

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.07 

Confusion Matrix: 

         observed

predicted    1    0

        1  187  361

        0   17 4436

AUC                          :  0.9709 

Percent Correctly Classified :  92.44151 

Sensitivity                  :  0.9166667 

Specificity                  :  0.9247446 

Kappa                        :  0.4655674 

True Skill Statistic         :  0.8414113 

   Calibration Statistics 

Intercept (general calibration)                            :  0.0009238045 

Slope   (direction and variation in fit)                   :  1.001183 

Testa0b1 (overall reliability of predictors)               :  0.9997946 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9962802 

Testb1|a (refinement given correct calibration)            :  0.9842613 

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

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.6178101  (sd 0.069725) 

NULL Deviance                : 0.34088 (sd 0.0062257)

Fit Deviance                 : 0.16037 (sd 0.017918)

Explained Deviance           : 0.18051 (sd 0.017729)

Percent Deviance Explained   : 52.958 (sd 5.1434)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.064 (sd 0.0069921) 

Confusion Matrix: 

         observed

predicted    1    0

        1  187  371

        0   17 4426

AUC                          :  0.96789  (sd 0.0090462) 

Percent Correctly Classified :  92.2417  (sd 0.96745) 

Sensitivity                  :  0.9171429  (sd 0.059728) 

Specificity                  :  0.9226631  (sd 0.01215) 

Kappa                        :  0.4598751  (sd 0.023377) 

True Skill Statistic         :  0.839806  (sd 0.050069) 

   Calibration Statistics 

Intercept (general calibration)                            :  -0.001208459  (sd 0.28397) 

Slope   (direction and variation in fit)                   :  1.013227  (sd 0.20238) 

Testa0b1 (overall reliability of predictors)               :  0.6321771  (sd 0.29036) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.6652521  (sd 0.33407) 

Testb1|a (refinement given correct calibration)            :  0.5590602  (sd 0.2517) 

Total time = 3 min

Boosted Regression Tree Modeling Results

Data:

G:\Results01\Withrow\Model\brt_1\CovariateCorrelationOutputMDS_initial.csv

n(pres)                 =      204

n(abs)                  =      4797

n covariates considered =      4

total time for model fitting = 0.25min

Settings:

random seed used             : 19680828

tree complexity              : 9

learning rate                : 0.0587

n(trees)                     : 300

model simplification         : cross-validation

n folds                      : 3

n covariates in final model  : 2

Relative influence of predictors in final model:

   Var  rel.inf

 bio16 53.08582

 bio10 46.91418

Important interactions in final model:

 v1 name1 v2 name2

  2 bio16  1 bio10

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

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.8236881 

NULL Deviance                : 0.34091

Fit Deviance                 : 0.098082

Explained Deviance           : 0.24282

Percent Deviance Explained   : 71.229

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.06 

Confusion Matrix: 

         observed

predicted    1    0

        1  192  279

        0   12 4518

AUC                          :  0.9878 

Percent Correctly Classified :  94.18116 

Sensitivity                  :  0.9411765 

Specificity                  :  0.9418386 

Kappa                        :  0.5428655 

True Skill Statistic         :  0.8830151 

   Calibration Statistics 

Intercept (general calibration)                            :  0.415043 

Slope   (direction and variation in fit)                   :  1.386246 

Testa0b1 (overall reliability of predictors)               :  3.16503e-07 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9916614 

Testb1|a (refinement given correct calibration)            :  4.475228e-08 

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

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.5560289  (sd 0.1004) 

NULL Deviance                : 0.34088 (sd 0.0062257)

Fit Deviance                 : 0.19923 (sd 0.042173)

Explained Deviance           : 0.14165 (sd 0.042982)

Percent Deviance Explained   : 41.524 (sd 12.418)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.085 (sd 0.015811) 

Confusion Matrix: 

         observed

predicted    1    0

        1  160  269

        0   44 4528

AUC                          :  0.95287  (sd 0.020284) 

Percent Correctly Classified :  93.74043  (sd 1.5999) 

Sensitivity                  :  0.7854762  (sd 0.10834) 

Specificity                  :  0.9439161  (sd 0.017084) 

Kappa                        :  0.4834514  (sd 0.092749) 

True Skill Statistic         :  0.7293923  (sd 0.10588) 

   Calibration Statistics 

Intercept (general calibration)                            :  -0.3314141  (sd 0.41688) 

Slope   (direction and variation in fit)                   :  0.7694473  (sd 0.13856) 

Testa0b1 (overall reliability of predictors)               :  0.1862941  (sd 0.26594) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.3955747  (sd 0.22739) 

Testb1|a (refinement given correct calibration)            :  0.2183448  (sd 0.38081) 

Total time = 3.84 min

MARS

Random Forests

MARS Model Results

Data:

G:\Results01\Withrow\Model\mars_1\CovariateCorrelationOutputMDS_initial.csv

n(pres)                 =      204

n(abs)                  =      4797

n covariates considered =      4

total time for model fitting = 0.06min

Settings:

random seed used             : 123

mars degree                  : 1

mars penalty                 : 2

Summary of Model:

             nsubsets   gcv    rss

bio16              13 100.0  100.0

bio10              12  52.1   53.6

bio02              11  43.8   45.5

di_all_model        2   7.2    9.3

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

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.658014 

NULL Deviance                : 0.34091

Fit Deviance                 : 0.15056

Explained Deviance           : 0.19034

Percent Deviance Explained   : 55.835

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.06 

Confusion Matrix: 

         observed

predicted    1    0

        1  188  357

        0   16 4440

AUC                          :  0.9693 

Percent Correctly Classified :  92.54149 

Sensitivity                  :  0.9215686 

Specificity                  :  0.9255785 

Kappa                        :  0.4705742 

True Skill Statistic         :  0.8471471 

   Calibration Statistics 

Intercept (general calibration)                            :  0.0007387854 

Slope   (direction and variation in fit)                   :  1.000987 

Testa0b1 (overall reliability of predictors)               :  0.99983 

Testa0|b1(incorrect calibration given correct refinement)  :  0.9961934 

Testb1|a (refinement given correct calibration)            :  0.9857871 

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

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.6353281  (sd 0.067819) 

NULL Deviance                : 0.34088 (sd 0.0062257)

Fit Deviance                 : 0.15851 (sd 0.017099)

Explained Deviance           : 0.18236 (sd 0.018973)

Percent Deviance Explained   : 53.473 (sd 5.1702)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.058 (sd 0.0042164) 

Confusion Matrix: 

         observed

predicted    1    0

        1  186  386

        0   18 4411

AUC                          :  0.96677  (sd 0.013424) 

Percent Correctly Classified :  91.92098  (sd 1.0347) 

Sensitivity                  :  0.9121429  (sd 0.054773) 

Specificity                  :  0.919532  (sd 0.01272) 

Kappa                        :  0.4477298  (sd 0.028567) 

True Skill Statistic         :  0.8316749  (sd 0.044412) 

   Calibration Statistics 

Intercept (general calibration)                            :  0.03508582  (sd 0.4758) 

Slope   (direction and variation in fit)                   :  1.02432  (sd 0.29119) 

Testa0b1 (overall reliability of predictors)               :  0.5164617  (sd 0.33284) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.6307558  (sd 0.34862) 

Testb1|a (refinement given correct calibration)            :  0.4756484  (sd 0.2746) 

Total time = 0.89 min

Random Forest Modeling Results

Data:

G:\Results01\Withrow\Model\rf_1\CovariateCorrelationOutputMDS_initial.csv

n(pres)                 =      204

n(abs)                  =      4797

n covariates considered =      4

total time for model fitting = 0.13min

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

bio16        22.6992 164.3124              78.8823          98.1769

bio10        38.0227  97.9466              59.9772          65.0626

bio02        50.8372  29.7127              59.8055          52.6246

di_all_model 17.9477  29.9477              36.9348          31.0002

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

Evaluation Statistics applied to train split:

Correlation Coefficient      : 0.643329 

NULL Deviance                : 0.34091

Fit Deviance                 : 0.15206

Explained Deviance           : 0.18885

Percent Deviance Explained   : 55.396

  Threshold Methods based on Sens=Spec 

Threshold                    :  0.05 

Confusion Matrix: 

         observed

predicted    1    0

        1  189  338

        0   15 4459

AUC                          :  0.9707 

Percent Correctly Classified :  92.94141 

Sensitivity                  :  0.9264706 

Specificity                  :  0.9295393 

Kappa                        :  0.4869226 

True Skill Statistic         :  0.8560099 

   Calibration Statistics 

Intercept (general calibration)                            :  -0.08168425 

Slope   (direction and variation in fit)                   :  0.9721286 

Testa0b1 (overall reliability of predictors)               :  0.7441919 

Testa0|b1(incorrect calibration given correct refinement)  :  0.6350263 

Testb1|a (refinement given correct calibration)            :  0.5454089 

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

Evaluation Statistics applied to crossValidation split:

Correlation Coefficient      : 0.6349022  (sd 0.079748) 

NULL Deviance                : 0.34088 (sd 0.0062257)

Fit Deviance                 : 0.16078 (sd 0.027534)

Explained Deviance           : 0.1801 (sd 0.028873)

Percent Deviance Explained   : 52.806 (sd 8.1273)

  Threshold Methods based on Sens=Spec 

Mean Threshold               : 0.4335 (sd 0.094753) 

Confusion Matrix: 

         observed

predicted    1    0

        1  108   81

        0   96 4716

AUC                          :  0.9645  (sd 0.019657) 

Percent Correctly Classified :  96.46047  (sd 0.62691) 

Sensitivity                  :  0.5292857  (sd 0.079262) 

Specificity                  :  0.9831133  (sd 0.0054287) 

Kappa                        :  0.531137  (sd 0.076738) 

True Skill Statistic         :  0.512399  (sd 0.079985) 

   Calibration Statistics 

Intercept (general calibration)                            :  -0.000168203  (sd 0.55347) 

Slope   (direction and variation in fit)                   :  0.9233272  (sd 0.26437) 

Testa0b1 (overall reliability of predictors)               :  0.3497528  (sd 0.34683) 

Testa0|b1(incorrect calibration given correct refinement)  :  0.5361855  (sd 0.29144) 

Testb1|a (refinement given correct calibration)            :  0.3331741  (sd 0.36097) 

Total time = 2 min