Question

我正在使用Logistic回归设置建立死亡率预测模型。为了选择变量，我通过MASS软件包的stepAIC函数对模型进行逐步回归。我使用该功能已经有一段时间了，没有太多麻烦。对于我当前的数据集，当我选择某些变量但为其他变量平稳执行时，看到此函数会引发错误“ x [good，，drop = FALSE]：（下标）逻辑下标过长”。

我将模型缩小为4个变量，但仍然可以看到结果出现。检查可变长度，最大可变长度约为14。我还检查了这些选定变量的变量分布，对我来说看起来不错。经检查的VIF，都在2-2.5左右。我什至删除了NA值。有趣的是，相同的代码对于具有其他变量集的相同数据集也可以顺利运行。任何帮助将不胜感激。

logitmodel = glm(death_count~ ., tempdata[,c("death_count", "pct50to74_acs", "pct75to99_acs", "pct100to124_acs")], 
                 offset = tempdata$offset_baseqx, family = "binomial")

nothing = glm(death_count ~ 1, tempdata[,c("death_count", "pct50to74_acs", "pct75to99_acs", "pct100to124_acs")], 
              offset = tempdata$offset_baseqx, family = "binomial")

myscope = list(lower = formula(nothing), upper = formula(logitmodel))

ForwardModel = stepAIC(nothing, scope = myscope, direction = "forward", k = 2)

下面是变量摘要。

> summary(tempdata)

  death_count      offset_baseqx       pct50to74_acs     pct75to99_acs     pct100to124_acs  
 Min.   :0.00000   Min.   :-12.68618   Min.   :0.00000   Min.   :0.00000   Min.   :0.00000  
 1st Qu.:0.00000   1st Qu.: -6.37140   1st Qu.:0.01375   1st Qu.:0.01812   1st Qu.:0.02183  
 Median :0.00000   Median : -5.13844   Median :0.02640   Median :0.03383   Median :0.03783  
 Mean   :0.01698   Mean   : -5.18591   Mean   :0.03229   Mean   :0.03784   Mean   :0.04040  
 3rd Qu.:0.00000   3rd Qu.: -3.95894   3rd Qu.:0.04486   3rd Qu.:0.05279   3rd Qu.:0.05511  
 Max.   :1.00000   Max.   :  0.03052   Max.   :0.10469   Max.   :0.10453   Max.   :0.10119

这是我面临的错误

> ForwardModel = stepAIC(nothing, scope = myscope, direction = "forward", k = 2)
Start:  AIC=224385.6
death_count ~ 1

Error in x[good, , drop = FALSE] : (subscript) logical subscript too long

StepAIC抛出错误，因为“ x [good，，drop = FALSE]中的错误：（下标）逻辑下标过长”

0 个答案: