Function auroc: how can i get the ROC plot data?

Analysis
1

Hello there

mixOmics did a great job in my study area. When i run the example, like:

data(breast.tumors)
X <- breast.tumors$gene.exp
Y <- breast.tumors$sample$treatment
splsda.breast <- splsda(X, Y, ncomp = 2, keepX = c(25, 25))
auc.splsda = auroc(splsda.breast, roc.comp = 2)

it will get ROC plot, like:

ROC
ROC748×669 33 KB

and i want to get the plot data, like:
auc.splsda$graph.Comp2$data

Well, i dont understand why the plot data just look like below, it doesn’t seem to correspond to the plot result. Actuall, when i use the data with more than 4 groups, the same problem was shown which the data doesnt correspond to the plot result. So, am i understand right about this problem? How can i get the ROC plot data?

                  Specificity Sensitivity      Outcome
(1.24, Inf]               100  100.000000 AF vs BE : 1
(1.16,1.24]                95  100.000000 AF vs BE : 1
(1.1,1.16]                 90  100.000000 AF vs BE : 1
(1.07,1.1]                 85  100.000000 AF vs BE : 1
(1.06,1.07]                80  100.000000 AF vs BE : 1
(1.05,1.06]                75  100.000000 AF vs BE : 1
(1.02,1.05]                70  100.000000 AF vs BE : 1
(0.977,1.02]               65  100.000000 AF vs BE : 1
(0.953,0.977]              60  100.000000 AF vs BE : 1
(0.919,0.953]              55  100.000000 AF vs BE : 1
(0.869,0.919]              50  100.000000 AF vs BE : 1
(0.845,0.869]              45  100.000000 AF vs BE : 1
(0.831,0.845]              40  100.000000 AF vs BE : 1
(0.818,0.831]              35  100.000000 AF vs BE : 1
(0.775,0.818]              30  100.000000 AF vs BE : 1
(0.727,0.775]              25  100.000000 AF vs BE : 1
(0.717,0.727]              20  100.000000 AF vs BE : 1
(0.687,0.717]              15  100.000000 AF vs BE : 1
(0.581,0.687]              10  100.000000 AF vs BE : 1
(0.427,0.581]               5  100.000000 AF vs BE : 1
(0.341,0.427]               0    0.000000 AF vs BE : 1
(0.302,0.341]               0    3.703704 AF vs BE : 1
(0.269,0.302]               0    7.407407 AF vs BE : 1
(0.24,0.269]                0   11.111111 AF vs BE : 1
(0.2,0.24]                  0   14.814815 AF vs BE : 1
(0.184,0.2]                 0   18.518519 AF vs BE : 1
(0.18,0.184]                0   22.222222 AF vs BE : 1
(0.176,0.18]                0   25.925926 AF vs BE : 1
(0.166,0.176]               0   29.629630 AF vs BE : 1
(0.126,0.166]               0   33.333333 AF vs BE : 1
(0.0923,0.126]              0   37.037037 AF vs BE : 1
(0.0869,0.0923]             0   40.740741 AF vs BE : 1
(0.082,0.0869]              0   44.444444 AF vs BE : 1
(0.0714,0.082]              0   48.148148 AF vs BE : 1
(0.0588,0.0714]             0   51.851852 AF vs BE : 1
(0.0518,0.0588]             0   55.555556 AF vs BE : 1
(0.0372,0.0518]             0   59.259259 AF vs BE : 1
(0.00668,0.0372]            0   62.962963 AF vs BE : 1
(-0.0319,0.00668]           0   66.666667 AF vs BE : 1
(-0.0604,-0.0319]           0   70.370370 AF vs BE : 1
(-0.0722,-0.0604]           0   74.074074 AF vs BE : 1
(-0.0955,-0.0722]           0   77.777778 AF vs BE : 1
(-0.131,-0.0955]            0   81.481481 AF vs BE : 1
(-0.175,-0.131]             0   85.185185 AF vs BE : 1
(-0.217,-0.175]             0   88.888889 AF vs BE : 1
(-0.239,-0.217]             0   92.592593 AF vs BE : 1
(-Inf,-0.239]               0   96.296296 AF vs BE : 1
                            0  100.000000 AF vs BE : 1
2

Hi @huwanjin,

The auc.splsda$graph.Comp2 object is a ggplot object which is plotted when you print it. The $data entry in the object simply reflects data created and used by ggplot to visualise the output, not the data used or generated by the model.

The auroc function simply uses the splsda model to predict the original data and generates the AUC plots for each component based on these predictions.

Hope it helps,

Al

3

Hi:

Thanks for your reply, i think i find the reason why i confused about my question.

The auroc function created the ROC plot, right? Back to the question I asked at the beginning. The label of the x coordinate of the ROC plot is 100-Specificity(%). But when using the script auc.splsda$graph.Comp2$data to show the plot data, i found that the header of the second column of data is Specificity, actually the correct header should be 100-Specificity(%), you can check the part of plot data:

(0.341,0.427]               0    0.000000 AF vs BE : 1
(0.302,0.341]               0    3.703704 AF vs BE : 1
(0.269,0.302]               0    7.407407 AF vs BE : 1
(0.24,0.269]                0   11.111111 AF vs BE : 1
(0.2,0.24]                  0   14.814815 AF vs BE : 1
(0.184,0.2]                 0   18.518519 AF vs BE : 1
(0.18,0.184]                0   22.222222 AF vs BE : 1
(0.176,0.18]                0   25.925926 AF vs BE : 1
(0.166,0.176]               0   29.629630 AF vs BE : 1
(0.126,0.166]               0   33.333333 AF vs BE : 1
(0.0923,0.126]              0   37.037037 AF vs BE : 1
(0.0869,0.0923]             0   40.740741 AF vs BE : 1
(0.082,0.0869]              0   44.444444 AF vs BE : 1
(0.0714,0.082]              0   48.148148 AF vs BE : 1
(0.0588,0.0714]             0   51.851852 AF vs BE : 1
(0.0518,0.0588]             0   55.555556 AF vs BE : 1
(0.0372,0.0518]             0   59.259259 AF vs BE : 1
(0.00668,0.0372]            0   62.962963 AF vs BE : 1
(-0.0319,0.00668]           0   66.666667 AF vs BE : 1
(-0.0604,-0.0319]           0   70.370370 AF vs BE : 1
(-0.0722,-0.0604]           0   74.074074 AF vs BE : 1
(-0.0955,-0.0722]           0   77.777778 AF vs BE : 1
(-0.131,-0.0955]            0   81.481481 AF vs BE : 1
(-0.175,-0.131]             0   85.185185 AF vs BE : 1
(-0.217,-0.175]             0   88.888889 AF vs BE : 1
(-0.239,-0.217]             0   92.592593 AF vs BE : 1
(-Inf,-0.239]               0   96.296296 AF vs BE : 1
                            0  100.000000 AF vs BE : 1

The second column of data should definitely be x coordinate of the ROC plot, which is 100-Specificity(%).

Although it is a small problem, it is easy to make users get confused. Maybe there are some ways to solve this problem?

I hope I clarified my question.

Thanks