How to compare two unpaired ROC curves
How to compare to unpaired ROC curves
This function was written to compare two unpaired ROC curves. What this means? This means that you have used the same classificator (i.e. a clinical test) on two different subsets of subjects. This function recalls another function of mine, ROC, to perform all the required computations.
The inputs are, as in ROC, two Nx2 matrix: in the first column you must insert the test value and in the second you must insert 1 if the subject is a patient or 0 if he is healty (or, more general, use 1 and 0 to discriminate the subsets of subjects).
I.E. load uroccompdata
and then call uroccomp(x,y)
The output is a plot:
This is an example ot uroccomp.m output plot
Then the function outputs the statistics computation:
ROC CURVES COMPARE
——————————————————————————–
………..ROC1 ROC2
——————————————————————————–
AUC 0.8994 0.9709
S.E. 0.0308 0.0166
——————————————————————————–
z 2-tails p-value
2.0445 0.040907 The areas are statistically different
where AUC is Area under the curve and S.E. is Standard Error.
If any problems occurs in execution, or if you found a bug, have a suggestion or question just contact me at:
giuseppe dot cardillo-edta at poste dot it
You can visit my homepage http://home.tele2.it/cardillo
My profile on XING http://www.xing.com/go/invita/13675097
My profile on LinkedIN http://it.linkedin.com/in/giuseppecardillo
Download Now
Popularity: 3% [?]
ROC curve
ROC – Receiver Operating Characteristics.
The ROC graphs are a useful technique for organizing classifiers and visualizing their performance. ROC graphs are commonly used in medical decision making.
The function computes and plots the classical ROC curve and the mirrored ROC curve (in my opinion, it is more useful).
An example of roc.m output plot
The input is a Nx2 matrix: in the first column you must insert the test value and in the second you must insert 1 if the subject is a patient or 0 if he is healthy (or, more general, use 1 and 0 to discriminate the subsets of subjects).
I.E. X=[165 1; 140 1; .... 166 0; 176 0] (load rocdata) then call roc(X).
The function will return a table
ROC CURVE DATA
——————————————–
Cut-off point Sensitivity Specificity
…
131.0000 0.2414 0.9762
132.0000 0.2586 0.9762
…
151.0000 0.7500 0.8095
152.0000 0.8017 0.8095
…
218.0000 1.0000 0.0476
239.0000 1.0000 0.0238
——————————————————————————–
and the statistic results:
ROC CURVE ANALYSIS
——————————————————————————–
AUC S.E. 95% C.I. Comment
——————————————————————————–
0.87541 0.02713 0.82224 0.92858 Good test
——————————————————————————–
Standardized AUC 1-tail p-value
13.8375 0.000000 The area is statistically greater than 0.5
Cut-off point for best Sensitivity and Specificity (blu circle in plot)= 152.0000
In the ROC plot, the cut-off point is the closest to [0,1] point or, if you want, the closest to the green line
where AUC is Area under the curve and S.E. is Standard Error
If you have downloaded partest (http://www.advancedmcode.org/partest.html) the routine will compute several data on test performance.
If any problems occurs in execution, or if you found a bug, have a suggestion or question just contact me at:
giuseppe dot cardillo-edta at poste dot it
You can visit my homepage http://home.tele2.it/cardillo
My profile on XING http://www.xing.com/go/invita/13675097
My profile on LinkedIN http://it.linkedin.com/in/giuseppecardillo
Download Now
Popularity: 5% [?]
Translator
Follow Us
Video Posts
