Question

我正在运行以下形式的逻辑回归：

model_1 <- lrm(dependent_variable ~ var1 + var2 + var3, data = merged_dataset, na.action="na.delete")

我想做的是将上一季度的平均因变量作为自变量。

我的数据如下：

> print(head(merged_dataset[,c("dependent_variable", "qy")]))
   dependent_variable      qy
1:                  0 2008 Q1
2:                  0 2008 Q1
3:                  0 2008 Q1
4:                  0 2008 Q1
5:                  0 2008 Q1
6:                  0 2008 Q1

其中因变量的取值为0或1，而qy变量取的值为诸如2008 Q1、2008 Q2，...，2017 Q4。

编辑：

我运行以下代码：

# function that gets the name of previous quarter
PrevQ = function(quarter = "2018 Q1"){
  ifelse(grepl("Q1", quarter), 
         paste0(as.numeric(substr(quarter, 1,4))-1, " Q4"), 
         paste0(substr(quarter, 1,6), as.numeric(substr(quarter,7,7))-1)
  )
}

# means in current quarters
merged_dataset$dep_means = with(merged_dataset, ave(dependent_variable, qy, FUN = mean))

# get names of previous quarters   
merged_dataset$prev_qy = PrevQ(merged_dataset$qy)

# merge mean of previous quarter by name of the quarter
merged_dataset$var4 = with(merged_dataset, dep_means[match(prev_qy, qy)])

print(head(merged_dataset[,c("dependent_variable", "qy")]))


print(head(cbind(merged_dataset$var4,merged_dataset$qy)))

这给了我

> print(head(merged_dataset[,c("dependent_variable", "qy")]))
   dependent_variable      qy
1:                  0 2008 Q1
2:                  0 2008 Q1
3:                  0 2008 Q1
4:                  0 2008 Q1
5:                  0 2008 Q1
6:                  0 2008 Q1
> 
> 
> print(head(cbind(merged_dataset$var4,merged_dataset$qy)))
     [,1] [,2]
[1,]   NA 2008
[2,]   NA 2008
[3,]   NA 2008
[4,]   NA 2008
[5,]   NA 2008
[6,]   NA 2008

似乎正在改变qy变量，仅保留年份？我得到了所有NA。

> print(head(cbind(merged_dataset$var4,merged_dataset$qy,merged_dataset$dep_means)))
     [,1] [,2]      [,3]
[1,]   NA 2008 0.1292719
[2,]   NA 2008 0.1292719
[3,]   NA 2008 0.1292719
[4,]   NA 2008 0.1292719
[5,]   NA 2008 0.1292719
[6,]   NA 2008 0.1292719


> print(tail(cbind(merged_dataset$var4,merged_dataset$qy,merged_dataset$dep_means)))
         [,1]    [,2]       [,3]
[32008,]   NA 2017.75 0.09802372
[32009,]   NA 2017.75 0.09802372
[32010,]   NA 2017.75 0.09802372
[32011,]   NA 2017.75 0.09802372
[32012,]   NA 2017.75 0.09802372
[32013,]   NA 2017.75 0.09802372

Answer 1

# function that gets the name of previous quarter
PrevQ = function(quarter = "2018 Q1"){
    ifelse(grepl("Q1", quarter), 
           paste0(as.numeric(substr(quarter, 1,4))-1, " Q4"), 
           paste0(substr(quarter, 1,6), as.numeric(substr(quarter,7,7))-1)
       )
}

# sample data
set.seed(5551)
merged_dataset = data.frame(dependent_variable = sample.int(2,6, rep = TRUE)-1, 
                            qy = rep(c("2017 Q4", "2018 Q1"), each = 3))

# means in current quarters
merged_dataset$dep_means = with(merged_dataset, ave(dependent_variable, qy, FUN = mean))

# get names of previous quarters   
merged_dataset$prev_qy = PrevQ(merged_dataset$qy)

# merge mean of previous quarter by name of the quarter
merged_dataset$var4 = with(merged_dataset, dep_means[match(prev_qy, qy)])

merged_dataset
#   dependent_variable      qy  dep_means prev_qy       var4
# 1                  0 2017 Q4 0.33333333 2017 Q3         NA
# 2                  1 2017 Q4 0.33333333 2017 Q3         NA
# 3                  0 2017 Q4 0.33333333 2017 Q3         NA
# 4                  0 2018 Q1 0.00000000 2017 Q4 0.33333333
# 5                  0 2018 Q1 0.00000000 2017 Q4 0.33333333
# 6                  0 2018 Q1 0.00000000 2017 Q4 0.33333333

R-计算上一季度因变量的平均值作为自变量

1 个答案: