医学统计python之ROC比较：Delong test

最新推荐文章于 2024-02-28 21:28:28 发布

原创最新推荐文章于 2024-02-28 21:28:28 发布 · 5.2k 阅读

12 ·

CC 4.0 BY-SA版权

文章标签：

#python

该代码使用R语言的pROC包进行ROC曲线的绘制与比较，通过'roc.test'函数以'delong'、'bootstrap'或'venakraman'方法计算两个ROC曲线的显著性差异，并返回p值。

def roc_test_r(targets_1, scores_1, targets_2, scores_2, method='delong'):
    # method: “delong”, “bootstrap” or “venkatraman”
    importr('pROC')
    robj.globalenv['targets_1'] = targets_1 = robj.FloatVector(targets_1)
    robj.globalenv['scores_1'] = scores_1 = robj.FloatVector(scores_1)
    robj.globalenv['targets_2'] = targets_2 = robj.FloatVector(targets_2)
    robj.globalenv['scores_2'] = scores_2 = robj.FloatVector(scores_2)

    r('roc_1 <- roc(targets_1, scores_1)')
    r('roc_2 <- roc(targets_2, scores_2)')
    r('result = roc.test(roc_1, roc_2, method="%s")' % method)
    p_value = r('p_value = result$p.value')
    return np.array(p_value)[0]