q****3
发帖数: 20 | 1 Two diagnostic tests:
Test A, only output a binary parameter, yes/no.
Test B, output binary parameters and continuous parameters.
Use measurements from gold standard as outcomes.
Thus, For A, we only have a single point in ROC space; For B, we have a
curve in ROC space.
Question: How to compare A to B, using ROC analysis? In other words, how to
compare a discrete classifier to a ROC curve?
I researched this topic for several days but didn't come up with solutions.
跪求大牛help!! | S******y
发帖数: 1123 | 2 Use confusion matrix instead
to
★ 发自iPhone App: ChineseWeb 11
【在 q****3 的大作中提到】
: Two diagnostic tests:
: Test A, only output a binary parameter, yes/no.
: Test B, output binary parameters and continuous parameters.
: Use measurements from gold standard as outcomes.
: Thus, For A, we only have a single point in ROC space; For B, we have a
: curve in ROC space.
: Question: How to compare A to B, using ROC analysis? In other words, how to
: compare a discrete classifier to a ROC curve?
: I researched this topic for several days but didn't come up with solutions.
: 跪求大牛help!!
| q****3
发帖数: 20 | 3 Different cut-off values will lead to different confusion matrix.
We don't know the optimization criteria for test A. How to compare test A
and test B if the optimization criteria are different? | t*****a
发帖数: 459 | 4 Test B, output binary parameters and continuous parameters.
这是啥意思？这两个output是分开的吗？还是说由continuous parameter上加一个
cutoff，给断成binary？ | q****3
发帖数: 20 | 5 It means, Test B output measurements which include binary/categorical
variables and continuous variables. | y**3
发帖数: 267 | | q****3
发帖数: 20 | 7 For binary outcome, c statistics=AUC. How to calculate AUC with only one
data point in the ROC space? | g******2
发帖数: 234 | 8 Control for recall in Test B (to match Test A), and compare precision, or
control for precision (to match Test A), and then compare recall. | y**3
发帖数: 267 | 9 Not sure what you are talking about. You mentioned binary parameter and
binary outcome.
My guess is your dependent variable is binary. For your A, the predictor set
are all binary; your B, the predictor set includes binary and continuous.
for either of the two, you can compute AUC for comparison
【在 q****3 的大作中提到】
: For binary outcome, c statistics=AUC. How to calculate AUC with only one
: data point in the ROC space?
| w*******n
发帖数: 469 | |
|
