In [1]:

    

options(warn=-1)

suppressMessages(library(dplyr))
source("loadData.R")

# load data
"Raw data:"
raw <- loadData()
format(head(raw,3), digits=3)

# clean data
"Cleaned data:"
clean <- cleanData(raw)
format(head(clean,3), digits=3)

# center and scale select features
"Centered and scaled data:"
wells <- unique(clean$Well.Name)
cs <- mutate(group_by(clean, Well.Name), 
             GR=as.numeric(scale(GR)), 
             ILD_log10=as.numeric(scale(ILD_log10)), 
             DeltaPHI=as.numeric(scale(DeltaPHI)), 
             PHIND=as.numeric(scale(PHIND)), 
             PE=as.numeric(scale(PE)))

format(head(cs,3), digits=3)


    

    





'Raw data:'






    






Facies
Formation
Well.Name
Depth
GR
ILD_log10
DeltaPHI
PHIND
PE
NM_M
RELPOS

	
3        
A1 SH    
SHRIMPLIN
2793     
77.5     
0.664    
 9.9     
11.9     
4.6      
1        
1.000    
	
3        
A1 SH    
SHRIMPLIN
2794     
78.3     
0.661    
14.2     
12.6     
4.1      
1        
0.979    
	
3        
A1 SH    
SHRIMPLIN
2794     
79.0     
0.658    
14.8     
13.1     
3.6      
1        
0.957    









    





'Cleaned data:'






    






Facies
Formation
Well.Name
Depth
GR
ILD_log10
DeltaPHI
PHIND
PE
isMarine
RELPOS

	
FSiS     
A1 SH    
SHRIMPLIN
2793     
77.5     
0.664    
 9.9     
11.9     
4.6      
FALSE    
1.000    
	
FSiS     
A1 SH    
SHRIMPLIN
2794     
78.3     
0.661    
14.2     
12.6     
4.1      
FALSE    
0.979    
	
FSiS     
A1 SH    
SHRIMPLIN
2794     
79.0     
0.658    
14.8     
13.1     
3.6      
FALSE    
0.957    









    





'Centered and scaled data:'






    






Facies
Formation
Well.Name
Depth
GR
ILD_log10
DeltaPHI
PHIND
PE
isMarine
RELPOS

	
FSiS     
A1 SH    
SHRIMPLIN
2793     
0.216    
 0.01855 
0.512    
-0.0487  
 0.421   
FALSE    
1.000    
	
FSiS     
A1 SH    
SHRIMPLIN
2794     
0.237    
 0.00567 
1.517    
 0.0736  
-0.133   
FALSE    
0.979    
	
FSiS     
A1 SH    
SHRIMPLIN
2794     
0.258    
-0.00721 
1.657    
 0.1648  
-0.687   
FALSE    
0.957    

In [29]:

    

table(cs$Facies[cs$isMarine])
table(cs$Facies[!cs$isMarine])


    

    






  SS CSiS FSiS SiSh   MS   WS    D   PS   BS 
   0    6   16  266  285  579  140  674  185 






    






  SS CSiS FSiS SiSh   MS   WS    D   PS   BS 
 268  934  764    5   11    3    1   12    0 

In [54]:

    

library(randomForest)
library(caret)
library(dplyr)
source("accuracyMetrics.R")

df <- cs[complete.cases(cs) & cs$Well.Name != "Recruit F9",]
formations <- unique(df$Formation)

fits <- list()

for (f in formations) {
    df_i <- df[df$Formation == f,]
    df_i$Facies <- droplevels(df_i$Facies)
    
    wells <- unique(df_i$Well.Name)
    resamp_list <- list()
    for (w in wells) {
        if (sum(df_i$Well.Name == w) > 0) {
            resamp_list[[w]] <- which(df_i$Well.Name != w)
        }
    }
    
    print(f)
    print(table(df_i$Well.Name))
    print(subset(df_i, select=-c(Well.Name)))
    
    fitControl <- trainControl(method="cv", index=resamp_list, summaryFunction=myF1MetricCaret)
    fits[[as.character(f)]] <- train(Facies ~ ., data=subset(df_i, select=-c(Well.Name)), method="rf",
                       trControl=fitControl, metric="F1")
    print(fits[[as.character(f)]])
    print(fits[[as.character(f)]][["resample"]])
    print("-----------------------------------------------------------------------------")
}


    

    




[1] "A1 SH"

    ALEXANDER D CHURCHMAN BIBLE  CROSS H CATTLE        KIMZEY A        LUKE G U 
              0              19               0               0              53 
          NEWBY           NOLAN      Recruit F9         SHANKLE       SHRIMPLIN 
             43              44               0              63              43 
Source: local data frame [265 x 10]

   Facies Formation  Depth        GR    ILD_log10  DeltaPHI       PHIND
   (fctr)    (fctr)  (dbl)     (dbl)        (dbl)     (dbl)       (dbl)
1    FSiS     A1 SH 2793.0 0.2155819  0.018551798 0.5120999 -0.04865478
2    FSiS     A1 SH 2793.5 0.2372980  0.005670378 1.5169013  0.07359111
3    FSiS     A1 SH 2794.0 0.2584779 -0.007211043 1.6571062  0.16480534
4    FSiS     A1 SH 2794.5 0.4474882 -0.020092464 1.4467989  0.17702993
5    FSiS     A1 SH 2795.0 0.1386372 -0.054442918 1.3533290  0.21182299
6    FSiS     A1 SH 2795.5 0.1222831 -0.101674794 1.4701664  0.22780899
7    FSiS     A1 SH 2796.0 0.1155806 -0.127437635 1.8440460  0.33030747
8    FSiS     A1 SH 2796.5 0.1673239 -0.148906669 2.0543532  0.32842676
9    FSiS     A1 SH 2797.0 0.1174573 -0.153200476 1.9842508  0.33971100
10   FSiS     A1 SH 2797.5 0.2005683 -0.191844738 2.1478231  0.38484794
..    ...       ...    ...       ...          ...       ...         ...
Variables not shown: PE (dbl), isMarine (lgl), RELPOS (dbl)
Random Forest 

265 samples
  9 predictor
  4 classes: 'SS', 'CSiS', 'FSiS', 'MS' 

No pre-processing
Resampling: Cross-Validated (10 fold) 
Summary of sample sizes: 222, 202, 212, 221, 222, 246, ... 
Resampling results across tuning parameters:

  mtry  F1       
   2    0.5389086
  11    0.5839428
  21    0.5823181

F1 was used to select the optimal model using  the largest value.
The final value used for the model was mtry = 11. 
         F1        Resample
1 0.3174603         SHANKLE
2 0.7027027       SHRIMPLIN
3 0.8181818           NOLAN
4 0.6800000        LUKE G U
5 0.4736842 CHURCHMAN BIBLE
6 0.5116279           NEWBY
[1] "-----------------------------------------------------------------------------"
[1] "A1 LM"

    ALEXANDER D CHURCHMAN BIBLE  CROSS H CATTLE        KIMZEY A        LUKE G U 
              0              94               0               0              67 
          NEWBY           NOLAN      Recruit F9         SHANKLE       SHRIMPLIN 
             82              61               0              40              51 
Source: local data frame [395 x 10]

   Facies Formation  Depth         GR  ILD_log10    DeltaPHI       PHIND
   (fctr)    (fctr)  (dbl)      (dbl)      (dbl)       (dbl)       (dbl)
1      PS     A1 LM 2814.5 -0.3321461 -0.1360252  0.90934700 -0.01574243
2      PS     A1 LM 2815.0 -0.7002472  0.1817165 -0.09545440 -1.02850258
3      PS     A1 LM 2815.5 -0.8940834  0.4393449 -0.46933399 -1.42533153
4      PS     A1 LM 2816.0 -0.6275922  0.7012671 -0.60953883 -1.42062976
5      WS     A1 LM 2816.5 -0.2442094  1.0962973 -0.72637620 -1.34728223
6      WS     A1 LM 2817.0  0.1557956  1.4569771 -0.77311115 -1.29368211
7      WS     A1 LM 2817.5  0.4767111  1.6544922 -0.70300873 -1.28898035
8      WS     A1 LM 2818.0  0.4678638  1.7360745 -0.58617136 -1.20152752
9      WS     A1 LM 2818.5  0.4984272  1.6802551 -0.28239419 -1.08962552
10     WS     A1 LM 2819.0  0.1624981  1.5385594  0.02138298 -0.93258657
..    ...       ...    ...        ...        ...         ...         ...
Variables not shown: PE (dbl), isMarine (lgl), RELPOS (dbl)
Random Forest 

395 samples
  9 predictor
  6 classes: 'FSiS', 'SiSh', 'MS', 'WS', 'D', 'PS' 

No pre-processing
Resampling: Cross-Validated (10 fold) 
Summary of sample sizes: 344, 355, 328, 334, 313, 301, ... 
Resampling results across tuning parameters:

  mtry  F1       
   2    0.4552256
  11    0.5213630
  21    0.4873485

F1 was used to select the optimal model using  the largest value.
The final value used for the model was mtry = 11. 
         F1        Resample
1 0.7750000         SHANKLE
2 0.7000000       SHRIMPLIN
3 0.4166667           NOLAN
4 0.7121212        LUKE G U
5 0.0000000 CHURCHMAN BIBLE
6 0.5243902           NEWBY
[1] "-----------------------------------------------------------------------------"
[1] "B1 SH"

    ALEXANDER D CHURCHMAN BIBLE  CROSS H CATTLE        KIMZEY A        LUKE G U 
              0              22               0               0              39 
          NEWBY           NOLAN      Recruit F9         SHANKLE       SHRIMPLIN 
             32              32               0              49              38 
Source: local data frame [212 x 10]

   Facies Formation  Depth          GR   ILD_log10   DeltaPHI       PHIND
   (fctr)    (fctr)  (dbl)       (dbl)       (dbl)      (dbl)       (dbl)
1    FSiS     B1 SH 2840.0  1.02846063  0.22894833 -1.3806655  1.38632385
2    FSiS     B1 SH 2840.5  1.11157158  0.11301555 -0.3992316  2.72256542
3    FSiS     B1 SH 2841.0  0.98985425  0.02284561  0.2083228  2.77710589
4    FSiS     B1 SH 2841.5  0.49440572 -0.08449957  0.4419975  1.80478092
5    FSiS     B1 SH 2842.0  0.30968493 -0.25195803  0.3017927  0.71773289
6    CSiS     B1 SH 2842.5 -0.06619106 -0.39365366  0.8158771  0.07359111
7    CSiS     B1 SH 2843.0 -0.21257358 -0.42371031  1.0028169 -0.09097066
8    CSiS     B1 SH 2843.5 -0.29890173 -0.39365366  1.0962868  0.02093134
9    CSiS     B1 SH 2844.0 -0.48523113 -0.32065894  1.1663892 -0.08250749
10   CSiS     B1 SH 2844.5 -0.03991728 -0.32924656  1.2832266 -0.08344784
..    ...       ...    ...         ...         ...        ...         ...
Variables not shown: PE (dbl), isMarine (lgl), RELPOS (dbl)
Random Forest 

212 samples
  9 predictor
  3 classes: 'SS', 'CSiS', 'FSiS' 

No pre-processing
Resampling: Cross-Validated (10 fold) 
Summary of sample sizes: 174, 163, 173, 180, 180, 190, ... 
Resampling results across tuning parameters:

  mtry  F1       
   2    0.5854924
  11    0.5428910
  21    0.5476950

F1 was used to select the optimal model using  the largest value.
The final value used for the model was mtry = 2. 
         F1        Resample
1 0.0000000         SHANKLE
2 0.6842105       SHRIMPLIN
3 0.7179487        LUKE G U
4 0.9545455 CHURCHMAN BIBLE
5 0.5000000           NEWBY
6 0.6562500           NOLAN
[1] "-----------------------------------------------------------------------------"
[1] "B1 LM"

    ALEXANDER D CHURCHMAN BIBLE  CROSS H CATTLE        KIMZEY A        LUKE G U 
              0              31               0               0              23 
          NEWBY           NOLAN      Recruit F9         SHANKLE       SHRIMPLIN 
             39              20               0              10              18 
Source: local data frame [141 x 10]

   Facies Formation  Depth         GR   ILD_log10    DeltaPHI      PHIND
   (fctr)    (fctr)  (dbl)      (dbl)       (dbl)       (dbl)      (dbl)
1      MS     B1 LM 2859.0  0.4190697 -0.13173144 -0.30576167  2.7263268
2      MS     B1 LM 2859.5 -0.6227664  0.03572703 -0.79647863  0.8343367
3      MS     B1 LM 2860.0 -0.9351027  0.18601027 -0.65627378 -0.7435756
4      MS     B1 LM 2860.5 -0.9174081  0.36205635  0.04475045 -1.4037034
5      PS     B1 LM 2861.0 -1.1141934  0.61539095  0.02138298 -1.4920966
6      PS     B1 LM 2861.5 -1.2313530  1.08770971 -0.39923156 -1.4958580
7      PS     B1 LM 2862.0 -1.3133916  1.59437892 -0.63290631 -1.5776687
8      PS     B1 LM 2862.5 -1.3099063  2.09246051 -0.84321358 -1.6980339
9      PS     B1 LM 2863.0 -1.3249199  2.71506251 -0.70300873 -1.8767010
10     PS     B1 LM 2863.5 -1.2930160  3.20026268 -0.58617136 -1.9904837
..    ...       ...    ...        ...         ...         ...        ...
Variables not shown: PE (dbl), isMarine (lgl), RELPOS (dbl)
Random Forest 

141 samples
  9 predictor
  5 classes: 'SiSh', 'MS', 'WS', 'D', 'PS' 

No pre-processing
Resampling: Cross-Validated (10 fold) 
Summary of sample sizes: 123, 131, 118, 121, 102, 110, ... 
Resampling results across tuning parameters:

  mtry  F1       
   2    0.4281388
  11    0.3962963
  21    0.3754902

F1 was used to select the optimal model using  the largest value.
The final value used for the model was mtry = 2. 
         F1        Resample
1 0.5000000         SHANKLE
2 0.3888889       SHRIMPLIN
3 0.4347826        LUKE G U
4 0.6451613 CHURCHMAN BIBLE
5 0.0000000           NEWBY
6 0.6000000           NOLAN
[1] "-----------------------------------------------------------------------------"
[1] "B2 SH"

    ALEXANDER D CHURCHMAN BIBLE  CROSS H CATTLE        KIMZEY A        LUKE G U 
              0              13               0               0              21 
          NEWBY           NOLAN      Recruit F9         SHANKLE       SHRIMPLIN 
             10              24               0              30              28 
Source: local data frame [126 x 10]

   Facies Formation  Depth          GR    ILD_log10   DeltaPHI      PHIND
   (fctr)    (fctr)  (dbl)       (dbl)        (dbl)      (dbl)      (dbl)
1    FSiS     B2 SH 2868.0 -0.04983697  1.315281470 -0.8198461 -1.3247138
2    FSiS     B2 SH 2868.5  0.12281933  0.868725557 -0.3758641 -0.8150425
3    FSiS     B2 SH 2869.0  0.36679019  0.628272372  0.3952626 -0.1831253
4    FSiS     B2 SH 2869.5  0.74722388  0.443638677  0.6523048  0.3378303
5    CSiS     B2 SH 2870.0  0.69038671  0.233242141  0.4419975  0.1478790
6    CSiS     B2 SH 2870.5  0.28073014  0.005670378  0.2083228 -0.3288800
7    CSiS     B2 SH 2871.0  0.18689519 -0.247664227  0.2784252 -0.6373157
8    CSiS     B2 SH 2871.5  0.16812821 -0.423710308  0.8626121 -0.3514484
9    CSiS     B2 SH 2872.0  0.40700517 -0.526761673  1.3299615  0.2955144
10   FSiS     B2 SH 2872.5  0.91210529 -0.578287355  1.6103712  0.8437402
..    ...       ...    ...         ...          ...        ...        ...
Variables not shown: PE (dbl), isMarine (lgl), RELPOS (dbl)
Random Forest 

126 samples
  9 predictor
  4 classes: 'SS', 'CSiS', 'FSiS', 'D' 

No pre-processing
Resampling: Cross-Validated (10 fold) 
Summary of sample sizes: 98, 96, 105, 102, 116, 113, ... 
Resampling results across tuning parameters:

  mtry  F1       
   2    0.4674603
  11    0.4809524
  21    0.4809524

F1 was used to select the optimal model using  the largest value.
The final value used for the model was mtry = 11. 
         F1        Resample
1 0.9333333         SHANKLE
2 0.6428571       SHRIMPLIN
3 0.0000000           NOLAN
4 0.8095238        LUKE G U
5 0.0000000 CHURCHMAN BIBLE
6 0.5000000           NEWBY
[1] "-----------------------------------------------------------------------------"
[1] "B2 LM"

    ALEXANDER D CHURCHMAN BIBLE  CROSS H CATTLE        KIMZEY A        LUKE G U 
              0              29               0               0              20 
          NEWBY           NOLAN      Recruit F9         SHANKLE       SHRIMPLIN 
             26              21               0               7              16 
Source: local data frame [119 x 10]

   Facies Formation  Depth         GR   ILD_log10   DeltaPHI      PHIND
   (fctr)    (fctr)  (dbl)      (dbl)       (dbl)      (dbl)      (dbl)
1       D     B2 LM 2882.0 -0.5015852 -0.39794747 -1.4274004  0.7271364
2       D     B2 LM 2882.5 -1.0348358 -0.25195803 -1.3572980 -1.2128118
3       D     B2 LM 2883.0 -1.3956982 -0.01579866 -0.8198461 -1.7760832
4      PS     B2 LM 2883.5 -1.3903362  0.37493777 -0.7030087 -1.7394094
5      PS     B2 LM 2884.0 -1.3731778  0.83866891 -0.4927015 -1.6717040
6      PS     B2 LM 2884.5 -1.4029369  1.38398238 -0.5394364 -1.5391142
7      PS     B2 LM 2885.0 -1.4351089  1.92070824 -0.5160689 -1.4732895
8      PS     B2 LM 2885.5 -1.3144640  2.19121807 -0.6329063 -1.4159280
9      PS     B2 LM 2886.0 -1.1924785  2.20409949 -0.6095388 -1.3924192
10     PS     B2 LM 2886.5 -1.0948902  2.10534193 -0.7497437 -1.4008824
..    ...       ...    ...        ...         ...        ...        ...
Variables not shown: PE (dbl), isMarine (lgl), RELPOS (dbl)






    




Error in {: task 1 failed - "subscript out of bounds"
Traceback:

1. train(Facies ~ ., data = subset(df_i, select = -c(Well.Name)), 
 .     method = "rf", trControl = fitControl, metric = "F1")   # at line 28-29 of file <text>
2. train.formula(Facies ~ ., data = subset(df_i, select = -c(Well.Name)), 
 .     method = "rf", trControl = fitControl, metric = "F1")
3. train(x, y, weights = w, ...)
4. train.default(x, y, weights = w, ...)
5. nominalTrainWorkflow(x = x, y = y, wts = weights, info = trainInfo, 
 .     method = models, ppOpts = preProcess, ctrl = trControl, lev = classLevels, 
 .     ...)
6. foreach(iter = seq(along = resampleIndex), .combine = "c", .verbose = FALSE, 
 .     .packages = pkgs, .errorhandling = "stop") %:% foreach(parm = 1:nrow(info$loop), 
 .     .combine = "c", .verbose = FALSE, .packages = pkgs, .errorhandling = "stop") %op% 
 .     {
 .         testing <- FALSE
 .         if (!(length(ctrl$seeds) == 1 && is.na(ctrl$seeds))) 
 .             set.seed(ctrl$seeds[[iter]][parm])
 .         loadNamespace("caret")
 .         if (ctrl$verboseIter) 
 .             progress(printed[parm, , drop = FALSE], names(resampleIndex), 
 .                 iter)
 .         if (names(resampleIndex)[iter] != "AllData") {
 .             modelIndex <- resampleIndex[[iter]]
 .             holdoutIndex <- ctrl$indexOut[[iter]]
 .         }
 .         else {
 .             modelIndex <- 1:nrow(x)
 .             holdoutIndex <- modelIndex
 .         }
 .         if (testing) 
 .             cat("pre-model\n")
 .         if (!is.null(info$submodels[[parm]]) && nrow(info$submodels[[parm]]) > 
 .             0) {
 .             submod <- info$submodels[[parm]]
 .         }
 .         else submod <- NULL
 .         mod <- try(createModel(x = x[modelIndex, , drop = FALSE], 
 .             y = y[modelIndex], wts = wts[modelIndex], method = method, 
 .             tuneValue = info$loop[parm, , drop = FALSE], obsLevels = lev, 
 .             pp = ppp, classProbs = ctrl$classProbs, sampling = ctrl$sampling, 
 .             ...), silent = TRUE)
 .         if (testing) 
 .             print(mod)
 .         if (class(mod)[1] != "try-error") {
 .             predicted <- try(predictionFunction(method = method, 
 .                 modelFit = mod$fit, newdata = x[holdoutIndex, 
 .                   , drop = FALSE], preProc = mod$preProc, param = submod), 
 .                 silent = TRUE)
 .             if (class(predicted)[1] == "try-error") {
 .                 wrn <- paste(colnames(printed[parm, , drop = FALSE]), 
 .                   printed[parm, , drop = FALSE], sep = "=", collapse = ", ")
 .                 wrn <- paste("predictions failed for ", names(resampleIndex)[iter], 
 .                   ": ", wrn, " ", as.character(predicted), sep = "")
 .                 if (ctrl$verboseIter) 
 .                   cat(wrn, "\n")
 .                 warning(wrn)
 .                 rm(wrn)
 .                 nPred <- length(holdoutIndex)
 .                 if (!is.null(lev)) {
 .                   predicted <- rep("", nPred)
 .                   predicted[seq(along = predicted)] <- NA
 .                 }
 .                 else {
 .                   predicted <- rep(NA, nPred)
 .                 }
 .                 if (!is.null(submod)) {
 .                   tmp <- predicted
 .                   predicted <- vector(mode = "list", length = nrow(info$submodels[[parm]]) + 
 .                     1)
 .                   for (i in seq(along = predicted)) predicted[[i]] <- tmp
 .                   rm(tmp)
 .                 }
 .             }
 .         }
 .         else {
 .             wrn <- paste(colnames(printed[parm, , drop = FALSE]), 
 .                 printed[parm, , drop = FALSE], sep = "=", collapse = ", ")
 .             wrn <- paste("model fit failed for ", names(resampleIndex)[iter], 
 .                 ": ", wrn, " ", as.character(mod), sep = "")
 .             if (ctrl$verboseIter) 
 .                 cat(wrn, "\n")
 .             warning(wrn)
 .             rm(wrn)
 .             nPred <- length(holdoutIndex)
 .             if (!is.null(lev)) {
 .                 predicted <- rep("", nPred)
 .                 predicted[seq(along = predicted)] <- NA
 .             }
 .             else {
 .                 predicted <- rep(NA, nPred)
 .             }
 .             if (!is.null(submod)) {
 .                 tmp <- predicted
 .                 predicted <- vector(mode = "list", length = nrow(info$submodels[[parm]]) + 
 .                   1)
 .                 for (i in seq(along = predicted)) predicted[[i]] <- tmp
 .                 rm(tmp)
 .             }
 .         }
 .         if (testing) 
 .             print(head(predicted))
 .         if (ctrl$classProbs) {
 .             if (class(mod)[1] != "try-error") {
 .                 probValues <- probFunction(method = method, modelFit = mod$fit, 
 .                   newdata = x[holdoutIndex, , drop = FALSE], 
 .                   preProc = mod$preProc, param = submod)
 .             }
 .             else {
 .                 probValues <- as.data.frame(matrix(NA, nrow = nPred, 
 .                   ncol = length(lev)))
 .                 colnames(probValues) <- lev
 .                 if (!is.null(submod)) {
 .                   tmp <- probValues
 .                   probValues <- vector(mode = "list", length = nrow(info$submodels[[parm]]) + 
 .                     1)
 .                   for (i in seq(along = probValues)) probValues[[i]] <- tmp
 .                   rm(tmp)
 .                 }
 .             }
 .             if (testing) 
 .                 print(head(probValues))
 .         }
 .         if (is.numeric(y)) {
 .             if (is.logical(ctrl$predictionBounds) && any(ctrl$predictionBounds)) {
 .                 if (is.list(predicted)) {
 .                   predicted <- lapply(predicted, trimPredictions, 
 .                     mod_type = "Regression", bounds = ctrl$predictionBounds, 
 .                     limits = ctrl$yLimits)
 .                 }
 .                 else {
 .                   predicted <- trimPredictions(mod_type = "Regression", 
 .                     bounds = ctrl$predictionBounds, limits = ctrl$yLimit, 
 .                     pred = predicted)
 .                 }
 .             }
 .             else {
 .                 if (is.numeric(ctrl$predictionBounds) && any(!is.na(ctrl$predictionBounds))) {
 .                   if (is.list(predicted)) {
 .                     predicted <- lapply(predicted, trimPredictions, 
 .                       mod_type = "Regression", bounds = ctrl$predictionBounds, 
 .                       limits = ctrl$yLimits)
 .                   }
 .                   else {
 .                     predicted <- trimPredictions(mod_type = "Regression", 
 .                       bounds = ctrl$predictionBounds, limits = ctrl$yLimit, 
 .                       pred = predicted)
 .                   }
 .                 }
 .             }
 .         }
 .         if (!is.null(submod)) {
 .             allParam <- expandParameters(info$loop[parm, , drop = FALSE], 
 .                 info$submodels[[parm]])
 .             allParam <- allParam[complete.cases(allParam), , 
 .                 drop = FALSE]
 .             predicted <- lapply(predicted, function(x, y, wts, 
 .                 lv, rows) {
 .                 x <- outcome_conversion(x, lv = lev)
 .                 out <- data.frame(pred = x, obs = y, stringsAsFactors = FALSE)
 .                 if (!is.null(wts)) 
 .                   out$weights <- wts
 .                 out$rowIndex <- rows
 .                 out
 .             }, y = y[holdoutIndex], wts = wts[holdoutIndex], 
 .                 lv = lev, rows = holdoutIndex)
 .             if (testing) 
 .                 print(head(predicted))
 .             if (ctrl$classProbs) {
 .                 for (k in seq(along = predicted)) predicted[[k]] <- cbind(predicted[[k]], 
 .                   probValues[[k]])
 .             }
 .             if (keep_pred) {
 .                 tmpPred <- predicted
 .                 for (modIndex in seq(along = tmpPred)) {
 .                   tmpPred[[modIndex]]$rowIndex <- holdoutIndex
 .                   tmpPred[[modIndex]] <- merge(tmpPred[[modIndex]], 
 .                     allParam[modIndex, , drop = FALSE], all = TRUE)
 .                 }
 .                 tmpPred <- rbind.fill(tmpPred)
 .                 tmpPred$Resample <- names(resampleIndex)[iter]
 .             }
 .             else tmpPred <- NULL
 .             thisResample <- lapply(predicted, ctrl$summaryFunction, 
 .                 lev = lev, model = method)
 .             if (testing) 
 .                 print(head(thisResample))
 .             if (length(lev) > 1 && length(lev) <= 50) {
 .                 cells <- lapply(predicted, function(x) flatTable(x$pred, 
 .                   x$obs))
 .                 for (ind in seq(along = cells)) thisResample[[ind]] <- c(thisResample[[ind]], 
 .                   cells[[ind]])
 .             }
 .             thisResample <- do.call("rbind", thisResample)
 .             thisResample <- cbind(allParam, thisResample)
 .         }
 .         else {
 .             if (is.factor(y)) 
 .                 predicted <- outcome_conversion(predicted, lv = lev)
 .             tmp <- data.frame(pred = predicted, obs = y[holdoutIndex], 
 .                 stringsAsFactors = FALSE)
 .             names(tmp)[1] <- "pred"
 .             if (!is.null(wts)) 
 .                 tmp$weights <- wts[holdoutIndex]
 .             if (ctrl$classProbs) 
 .                 tmp <- cbind(tmp, probValues)
 .             tmp$rowIndex <- holdoutIndex
 .             if (keep_pred) {
 .                 tmpPred <- tmp
 .                 tmpPred$rowIndex <- holdoutIndex
 .                 tmpPred <- merge(tmpPred, info$loop[parm, , drop = FALSE], 
 .                   all = TRUE)
 .                 tmpPred$Resample <- names(resampleIndex)[iter]
 .             }
 .             else tmpPred <- NULL
 .             thisResample <- ctrl$summaryFunction(tmp, lev = lev, 
 .                 model = method)
 .             if (length(lev) > 1 && length(lev) <= 50) 
 .                 thisResample <- c(thisResample, flatTable(tmp$pred, 
 .                   tmp$obs))
 .             thisResample <- as.data.frame(t(thisResample))
 .             thisResample <- cbind(thisResample, info$loop[parm, 
 .                 , drop = FALSE])
 .         }
 .         thisResample$Resample <- names(resampleIndex)[iter]
 .         if (ctrl$verboseIter) 
 .             progress(printed[parm, , drop = FALSE], names(resampleIndex), 
 .                 iter, FALSE)
 .         if (testing) 
 .             print(thisResample)
 .         list(resamples = thisResample, pred = tmpPred)
 .     }
7. e$fun(obj, substitute(ex), parent.frame(), e$data)