Chapter 12 Analyzing a dataset with only diseased cases
12.2 The problem
How to analyze \(K_1 = 0\) datasets.
ROC-like plot of TPF vs. FPF1 is possible, see Section 3.12.1. Can create a ROC-like dataset with equal number of “non-diseased” and diseased cases (the ratings of the non-diseased cases are the FP ratings on diseased cases). Fit RSM to this dataset. Proceed as before. Key assumption being violated: the FP ratings on diseased cases are independent of the TP ratings on same cases. However, without this assumption one cannot estimate RSM parameters. Need RJafroc function to handle this special case: FitRsmRoc1? No! Just need function to create a “ROC” dataset from one that only has diseased cases. e.g., DfNoNormalsDataset?
12.2.1 Step 1: Create a test (diseased cases only) dataset
Save TONY dataset to dsTony. Create copy dsNoNormals. Remove all normal cases from it.
dsTony <- RJafroc::dataset01 # TONY dataset
K2 <- length(dsTony$lesions$perCase)
K1 <- length(dsTony$ratings$NL[1,1,,1]) - K2
dsNoNormals <- dsTony
# Remove all normal cases
dsNoNormals$ratings$NL <- dsNoNormals$ratings$NL[,,-(1:K1),]
# And fix truthTableStr
dsNoNormals$descriptions$truthTableStr <-
dsNoNormals$descriptions$truthTableStr[,,-(1:K1),]
RJafroc::UtilFigureOfMerit(dsTony,FOM = "wAFROC")
#> rdr1 rdr2 rdr3 rdr4 rdr5
#> trtBT 0.7602704 0.8406191 0.8171524 0.8153090 0.8278324
#> trtDM 0.6425854 0.7049977 0.7518434 0.7724426 0.6836962
#RJafroc::UtilFigureOfMerit(dsNoNormals,FOM = "wAFROC") #this will generate an error
RJafroc::UtilFigureOfMerit(dsTony,FOM = "wAFROC1")
#> rdr1 rdr2 rdr3 rdr4 rdr5
#> trtBT 0.8079866 0.8696629 0.8747798 0.8517613 0.8563468
#> trtDM 0.7277103 0.7781506 0.8225630 0.7968418 0.7496963
RJafroc::UtilFigureOfMerit(dsNoNormals,FOM = "wAFROC1")
#> rdr1 rdr2 rdr3 rdr4 rdr5
#> trtBT 0.8594559 0.9009910 0.9369398 0.8910807 0.8871039
#> trtDM 0.8195304 0.8570572 0.8988448 0.8231600 0.8208875
st <- St(dsTony,FOM = "wAFROC")
st1 <- St(dsNoNormals,FOM = "wAFROC1")
st$RRRC
#> $FTests
#> DF MS FStat p
#> Treatment 1.00000 0.025564954 10.29883 0.003668578
#> Error 24.70276 0.002482317 NA NA
#>
#> $ciDiffTrt
#> Estimate StdErr DF t PrGTt CILower
#> trtBT-trtDM 0.1011236 0.03151074 24.70276 3.209178 0.003668578 0.03618638
#> CIUpper
#> trtBT-trtDM 0.1660608
#>
#> $ciAvgRdrEachTrt
#> Estimate StdErr DF CILower CIUpper Cov2
#> trtBT 0.8122367 0.02698434 59.28149 0.7582465 0.8662268 0.0005390098
#> trtDM 0.7111131 0.03391021 17.78930 0.6398098 0.7824163 0.0006046324
st1$RRRC
#> $FTests
#> DF MS FStat p
#> Treatment 1.0000 0.0065582806 7.957961 0.005193632
#> Error 236.8821 0.0008241157 NA NA
#>
#> $ciDiffTrt
#> Estimate StdErr DF t PrGTt CILower
#> trtBT-trtDM 0.05121828 0.01815616 236.8821 2.820986 0.005193632 0.01545011
#> CIUpper
#> trtBT-trtDM 0.08698645
#>
#> $ciAvgRdrEachTrt
#> Estimate StdErr DF CILower CIUpper Cov2
#> trtBT 0.8951143 0.01974550 24.73302 0.8544254 0.9358031 0.0002330913
#> trtDM 0.8438960 0.02497063 27.62144 0.7927144 0.8950776 0.0003862498dsNoNormalsis the dataset with no non-diseased cases.stcontains the results of significance testing using the wAFROC-AUC figure of merit for the full dataset.st1contains the results of significance testing using the wAFROC1-AUC figure of merit for the dataset with no non-diseased cases.