R，Python或八度音阶:具有置信区间的经验分位数(反CDF)? [英] R, python or octave: empirical quantile (inverse cdf) with confidence intervals?

问题描述

我正在寻找一个内置函数，该函数可在除MATLAB之外的其他方式中返回样本分位数和估计的置信区间(MATLAB的ecdf会执行此操作).

I'm looking for a built-in function that returns the sample quantile and an estimated confidence interval in something other than MATLAB (MATLAB's ecdf does this).

我想R内置了这个功能，但我还没有找到它.

I'm guessing R has this built-in and I just haven't found it yet.

如果您有任何独立的代码可以执行此操作，虽然我希望找到包含在更大的开放代码库中的内容，但是您也可以在这里指向它.

If you have any standalone code to do this, you could also point to it here, though I hope to find something that is included as part of a larger open code base.

-试图摆脱MATLAB.

-Trying to get away from MATLAB.

推荐答案

survfit函数可用于获取具有置信区间的 survival 函数.由于它只是1-ecdf，因此分位数之间存在直接关系.要使用此功能，您必须创建一个变量，说明您的每个观察结果都是完整的(未经审查):

The survfit function can be used to get the survival function with confidence intervals. Since it is just 1-ecdf, there is a direct relationship between the quantiles. To use this you have to create a variable that says that each of your observations is complete (not censored):

library(survival)
x <- rexp(10)
ev <- rep(1, length(x))
sf <- survfit(Surv(x,ev)~1)

有输出:

>summary(sf)
Call: survfit(formula = Surv(x, ev) ~ 1)

     time n.risk n.event survival std.err lower 95% CI upper 95% CI
-1.4143     10       1      0.9  0.0949       0.7320        1.000
-1.1229      9       1      0.8  0.1265       0.5868        1.000
-0.9396      8       1      0.7  0.1449       0.4665        1.000
-0.4413      7       1      0.6  0.1549       0.3617        0.995
-0.2408      6       1      0.5  0.1581       0.2690        0.929
-0.1698      5       1      0.4  0.1549       0.1872        0.855
 0.0613      4       1      0.3  0.1449       0.1164        0.773
 0.1983      3       1      0.2  0.1265       0.0579        0.691
 0.5199      2       1      0.1  0.0949       0.0156        0.642
 0.8067      1       1      0.0     NaN           NA           NA

实际上，survfit确实计算了中位数及其置信区间，但没有计算其他分位数:

In fact, survfit does calculate the median and its confidence interval, but not the other quantiles:

>sf
Call: survfit(formula = Surv(x, ev) ~ 1)

records   n.max n.start  events  median 0.95LCL 0.95UCL 
 10.000  10.000  10.000  10.000  -0.205  -0.940      NA 

survival:::survmean函数很好地隐藏了计算中位数置信区间的实际工作，您可以使用它来推广到其他分位数.

The actual work for of the calculation of the confidence interval of the median is well hidden in the survival:::survmean function, which you could probably use to generalize to other quantiles.

这篇关于R，Python或八度音阶:具有置信区间的经验分位数(反CDF)?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！