R：如何为循环而下降的因素有<2级水平的多重线性回归

Question

我想回路多元线性回归和自动降不至少有两个层面，以避免以下错误消息因素：

误差在contrasts<-（*tmp*，值= contr.funs [1 + ISOF [NN]]）：对比度可以仅用于有2级或更多的水平的因素*

现在，我的代码是：

df %>% 
  group_by(crop_name) %>% 
    do(tidy(lm(formula = value ~ intercrop + 
erosion_c + purchased_seed + inorg_pest +
 org_pest + landscape + fert + inorgfert,
             data = . )))

问题是，一些农作物有足够的所有，我回归变量点的大样本，而其他有一个非常小的样本量和零收到一个给定的治疗（即，无血水果作物间作了，等等。 ）。

有没有一种办法中的for循环让R退步什么可以，否则放下一切，并避免这个问题？

Answer 1

我很新，所以这可能不是最好的办法。您可能需要建立一个for循环与crop_name，因为在我的例子DF是你的一个作物组子集。

df <- data.frame(intercrop = c("A","B","C","A","B","C"),
                   erosion_c = c("A","D","C","A","B","C"),
                   purchased_seed = c("A","B","D","F","E","C"),
                   inorg_pest = c("A","B","C","A","B","C"),
                   org_pest = c("A","B","A","A","B","B"),
                   landscape = c("A","A","A","A","A","A"),
                   fert = c("A","B","C","A","B","C"),
                   inorgfert = c("A","B","C","A","B","C")
                   )


yo <- sapply(df, levels)
hi <- as.data.frame(c(NA))
for(i in 1:length(yo)){
  hi[i] <- length(yo[[i]])
  names(hi)[i] <- names(df[i])
}

hi <- subset(as.data.frame(t(hi)), V1 >= 2)

formu <- row.names(hi)
formu <- as.formula(paste("value ~ ",gsub('.{3}$', '', paste( unlist(paste(formu,"+ ")), collapse=''))))