[英]Zero-inflated negative binomial distribution function NaN warning
I am trying to fit my data to a zero-inflated negative binomial model but one of my 3 independent variables (Exposure) seems to be causing NaNs to be produced (see very end of zeroinfl call) when the SE is being calculated in the summary function. I have also tried running a negative binomial hurdle model and am running into a similar issue.我正在尝试将我的数据拟合为零膨胀的负二项式 model 但当在摘要中计算 SE 时,我的 3 个自变量之一(曝光)似乎导致产生 NaN(请参阅 zeroinfl 调用的最后) function。我也尝试过运行负二项式障碍 model 并且遇到了类似的问题。
str(eggTreat)
'data.frame': 455 obs. of 4 variables:
$ Exposure : Factor w/ 2 levels "C","E": 2 2 2 2 2 2 2 2 2 2 ...
$ hi_lo : Factor w/ 2 levels "hi","lo": 2 2 2 2 2 2 2 2 2 2 ...
$ Egg_count: int 0 0 0 0 0 0 0 0 0 0 ...
$ Food : Factor w/ 2 levels "1.5A5YS","5ASMQ": 2 2 2 2 2 2 2 2 2 2 ...
mod.zeroinfl <- zeroinfl(Egg_count ~ Food+Exposure+hi_lo | Food+Exposure+hi_lo, data=eggTreat,
+ dist="negbin")
> summary(mod.zeroinfl)
Call:
zeroinfl(formula = Egg_count ~ Food + Exposure + hi_lo | Food + Exposure + hi_lo, data = eggTreat, dist = "negbin")
Pearson residuals:
Min 1Q Median 3Q Max
-0.65632 -0.47163 -0.28588 0.02976 9.00804
Count model coefficients (negbin with log link):
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.04435 0.14393 -0.308 0.7580
Food -1.12486 0.22267 -5.052 4.38e-07 ***
Exposure -2.34990 0.38684 -6.075 1.24e-09 ***
hi_lo -0.44893 0.19524 -2.299 0.0215 *
Log(theta) -0.24387 0.22639 -1.077 0.2814
Zero-inflation model coefficients (binomial with logit link):
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.830e+01 NA NA NA
Food -5.768e+00 5.628e+04 0 1
Exposure 4.612e-01 NA NA NA
hi_lo -7.477e+00 9.963e+05 0 1
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Theta = 0.7836
Number of iterations in BFGS optimization: 21
Log-likelihood: -350.2 on 9 Df
Warning message:
In sqrt(diag(object$vcov)) : NaNs produced
function (object, ...)
{
object$residuals <- residuals(object, type = "pearson")
kc <- length(object$coefficients$count)
kz <- length(object$coefficients$zero)
se <- sqrt(diag(object$vcov))
This problem is typically caused by complete separation ;这个问题通常是由完全分离引起的; using this search term, or searching for the Hauck-Donner effect, will show you that the problem is that there is some linear combination of your predictor values that perfectly separates the zeros and non-zeros (since the predictor variables in your zero-inflation are all categorical, this translates to a combination of categories where all the values are zero or non-zero).使用这个搜索词,或者搜索Hauck-Donner效应,会告诉你问题是你的预测值有一些线性组合可以完美地区分零和非零(因为你的零通货膨胀中的预测变量都是分类的,这转换为所有值为零或非零的类别组合)。
I would take a look at with(eggTreat, table(eggcount>0, Food, Exposure, hi_lo))
(arrange the arguments in whatever order makes the table easiest to read).我会看一下with(eggTreat, table(eggcount>0, Food, Exposure, hi_lo))
(以任何顺序排列 arguments,使表格最容易阅读)。
Typical symptoms include:典型症状包括:
|beta|>10
);较大的参数值(例如|beta|>10
); in this case your intercept is -18.3, which gives a predicted zero-inflation probability of 1e-8
in the baseline category (two of the other values are also large, although not nearly as extreme as the intercept)在这种情况下,您的截距是 -18.3,这给出了基线类别中1e-8
的预测零通胀概率(其他两个值也很大,尽管没有截距那么极端)Food
, hi_lo
), leading to z-values that are effectively zero and p-values of effectively 1极大的标准误差( Food
, hi_lo
),导致 z 值实际上为零,p 值实际上为 1NA
values you're seeing或...您看到的NA
值There are various solutions to this problem:这个问题有多种解决方案:
Zero-inflation model coefficients (binomial with logit link):
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.830e+01 NA NA NA
Food -5.768e+00 5.628e+04 0 1
Exposure 4.612e-01 NA NA NA
hi_lo -7.477e+00 9.963e+05 0 1
An answer to a similar problem has been posted here CrossValidated-NA-ZINB but what I found useful was to recalibrate my variables: eg.此处已发布类似问题的答案CrossValidated-NA-ZINB但我发现有用的是重新校准我的变量:例如。 I had the number of hectares of forest in a village that ranged 0 - 100,000 and turned them into hundreds of sqkm that ranged from 0 - 10 and the NaN that were shown for Std.我有一个村庄的森林公顷数,范围为 0 - 100,000,并将它们转换为数百平方公里,范围为 0 - 10 和标准显示的 NaN。 Error, z value and Pr(>|z|) are now valid numbers错误,z 值和 Pr(>|z|) 现在是有效数字
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.