应用函数来查找分布的高密度区域（编码） [英] Applying a function to find high density area of a distribution (coding)

问题描述

高密度区域分布在-r / 42987104＃42987104> 一个名为HDIofICDF的R函数 （也见下文）它为任何分布（单峰曲线）提供了两个极限值，从一个极限值到另一个极限值覆盖了该分布的高密度区域的95％。

该代码要求用户放置反向cdf 分布。对于R认可的分销，这可以通过 q分销名称来实现，例如， qf ， qchisq 等，以及，，然后 qdistribution name 。例如，对于F分布，可以通过以下方式找到两个极限值：
$ b $ HDIofICDF（qf，df1 = 10，df2 = 90） code> OR对于卡方分布，可以使用： HDIofICDF（qchisq，df = 10）


编码问题：

假设我已经创建了自己的发行版并拥有逆cdf 这个分配。我使用逆cdf （类似于R中的 qdistribution name ）（仅作为示例来显示我的<反向cdf 使用）：
$ b $ invcdf.posterior（p = .025，t = 2.81，N1 = 10，N2 = 10，rscale = 1）

现在：给出我的 invcdf.posterior 及其上面的参数，我如何使用 HDIofICDF 函数获得我的分布的两个限制值？

这是我的R代码，我有：

  HDIofICDF = function（ICDFname，credMass = 0.95， tol = 1e-8，...）{
 
 incredMass = 1.0  -  credMass 
 intervalWidth = function（lowTailPr，ICDFname，credMass，...）{
 ICDFname（credMass + lowTailPr，...） -  ICDFname（lowTailPr，...）
} 
 optInfo = optimize（intervalWidth，c（0，incredMass），ICDFname = ICDFname，
 credMass = credMass， tol = tol，...）
 HD IlowTailPr = optInfo $ minimum 
 return（c（ICDFname（HDIlowTailPr，...），
 ICDFname（credMass + HDIlowTailPr，...）））
} 
 ### ################################################## ##################### 
 ##使用示例1：## 
 
 HDIofICDF（qf，df1 = 10， df2 = 90）##这是一个F分配工作OK 
  

这是我如何尝试应用HDIofICDF函数对我自己的分发：
$ b $ pre $ HD $ f $ c HDIofICDF（invcdf.posterior，t = 2.81，N1 = 10，N2 = 10，rscale = 1）##不工作

我得到的错误是：

eval（expr，envir，enclos）中的错误：
参数t丢失，没有默认
调用从：eval（expr，envir，enclos）

解决方案

> HDIofICDF 函数有一个参数 tol 。

由于部分参数匹配，您调用 HDIofICDF（invcdf.posterior，t = 2.81，N1 = 10，N2 = 10，rscale = 1）， t 是指 tol 。 （你可以通过修改 HDIofICDF 函数来打印 tol ）来验证。）因此，当 invcdf.posterior 随后被调用，R抱怨缺少 t 参数。

下面是一个玩具示例，说明发生了什么：

  fun1 < - 函数（a，b）a + b 
 fun2 < - 函数（aaa，...）fun1（...）
 fun2（a = 1，b = 2）
＃fun1（...） ：参数a缺失，没有默认
 fun2（aaa = 99，a = 1，b = 2）
＃[1] 3 
  

如上例所示，为了解决您的问题，请在您的调用中明确指定 tol HDIofICDF ，例如

  HDIofICDF（invcdf.posterior，t = 2.81， N1 = 10，N2 = 10，rscale = 1，tol = 1e-8）
  

Background:

Recently, I came across an R function called HDIofICDF (also see below) that provides two limit-values for any distribution (a unimodal curve) such that from one limit-value to the other limit-value covers 95% highly dense area of that distribution.

The code requires that user put Inverse cdf of a distribution. For R-recognized distributions, this is achieved by "qdistribution name" e.g., qf, qchisq etc., and "," and then the arguments needed for the "qdistribution name". For example for an F distribution finding the two limit values is possible via:

HDIofICDF( qf , df1 = 10 , df2 = 90 ) OR for a Chi-Square Distribution, one can use: HDIofICDF( qchisq, df = 10)

Coding Question:

Suppose I have created my own distribution and have the Inverse cdf of this distribution. My Inverse cdf (similar to a "qdistribution name" in R) is given when I use (just as an example to show what arguments my Inverse cdf uses):

invcdf.posterior(p = .025, t = 2.81, N1 = 10, N2 = 10, rscale = 1 )

Now: given my invcdf.posterior and its arguments above, how can I use the HDIofICDF function to obtain the two limit-values for my distribution?

Here is my R codes I have:

HDIofICDF = function( ICDFname , credMass=0.95 , tol=1e-8 , ... ) {

 incredMass = 1.0 - credMass
 intervalWidth = function( lowTailPr , ICDFname , credMass , ... ) {
 ICDFname( credMass + lowTailPr , ... ) - ICDFname( lowTailPr , ... )
}
optInfo = optimize( intervalWidth , c( 0 , incredMass ) , ICDFname=ICDFname ,
                  credMass=credMass , tol=tol , ... )
 HDIlowTailPr = optInfo$minimum
 return( c( ICDFname( HDIlowTailPr , ... ) ,
         ICDFname( credMass + HDIlowTailPr , ... ) ) )
}
##########################################################################
## Example 1 of use: ##

HDIofICDF( qf , df1 = 10 , df2 = 90 ) ## This is a F distribution working OK

THIS IS HOW I TRIED TO APPLY THE HDIofICDF function TO MY OWN DISTRIBUTION:

HDIofICDF( invcdf.posterior, t = 2.81, N1 = 10 , N2 = 10, rscale = 1 ) ## Not working

The error I get is:

Error in eval(expr, envir, enclos) : argument "t" is missing, with no default Called from: eval(expr, envir, enclos)

解决方案

Notice that your HDIofICDF function has an argument tol.

Because of partial argument matching, when you call HDIofICDF(invcdf.posterior, t = 2.81, N1 = 10 , N2 = 10, rscale = 1), the t that you specified is assumed to mean tol. (You can verify this by modifying your HDIofICDF function to print tol.) Because of this, when your invcdf.posterior is subsequently called, R complains that the t argument is missing.

Here's a toy example illustrating what is going on:

fun1 <- function(a, b) a + b
fun2 <- function(aaa, ...) fun1(...)
fun2(a = 1, b = 2)
# Error in fun1(...) : argument "a" is missing, with no default
fun2(aaa = 99, a = 1, b = 2)
# [1] 3

As suggested by the example above, to get around your problem, explicitly specify tol in your call to HDIofICDF, e.g.

HDIofICDF( invcdf.posterior, t = 2.81, N1 = 10 , N2 = 10, rscale = 1, tol = 1e-8)

这篇关于应用函数来查找分布的高密度区域（编码）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！