[英]Correlation matrix in panel data in R
I have a time-series panel dataset which is structured in the following way: 我有一个时间序列面板数据集,其结构如下:
df <- data.frame(
year = c(2012L, 2013L, 2014L, 2012L, 2013L, 2014L),
id = c(1L, 1L, 1L, 2L, 2L, 2L),
c = c(11L, 13L, 13L, 16L, 15L, 15L)
)
#> year id c
#> 1 2012 1 11
#> 2 2013 1 13
#> 3 2014 1 13
#> 4 2012 2 16
#> 5 2013 2 15
#> 6 2014 2 15
I would like to find the cross-correlation between values in column C given their id number. 我想找到给定ID号的C列中的值之间的相互关系。 Something similar to this: 类似于以下内容:
#> 1 2
#> 1 1 0.8
#> 2 0.8 1
I have been using dplyr package to find the cross-correlation between two variables in my panel data but for some reason, I can't do the same for cross correlation in one veriable grouped by id. 我一直在使用dplyr包来查找面板数据中两个变量之间的互相关性,但是由于某些原因,对于一个按id分组的可验证的互相关性,我无法做到相同。
Do you mean something like the following? 您是说以下意思吗? I used the reshape package to cast based on the value of your id, followed by the cor()
function in baseR. 我使用了reshape包根据您的id的值进行了转换,随后是baseR中的cor()
函数。
> mydf <- data.frame(year=c("12","13","14","12","13","14"),id=c(1,1,1,2,2,2),c=c(11,13,13,16,15,156))
> library(reshape2)
> mydf
year id c
1 12 1 11
2 13 1 13
3 14 1 13
4 12 2 16
5 13 2 15
6 14 2 156
> my_wide_data <- dcast(mydf, year~id,value.var="c")
> cor(my_wide_data[,2:3])
1 2
1 1.0000000 0.4946525
2 0.4946525 1.0000000
So @Henrik's comment was much more simple and elegant, so including here. 因此,@ Henrik的评论更加简单和优雅,因此请在此处添加。
> cor(unstack(mydf[ , -1], c ~ id))
X1 X2
X1 1.0000000 0.4946525
X2 0.4946525 1.0000000
If you are already using tidyverse
tools, you should try widyr
. 如果您已经在使用tidyverse
工具,则应该尝试widyr
。
Its functions reshape to wide, get the correlations, and give you back a tidy data frame again. 它的功能可以调整为更宽的范围,获得相关性,并再次给您一个整洁的数据帧。
(Note I changed the sample data slightly to match akaDrHouse's answer. (请注意,我稍微更改了示例数据以匹配akaDrHouse的答案。
df <- data.frame(
year = c(2012L, 2013L, 2014L, 2012L, 2013L, 2014L),
id = c(1L, 1L, 1L, 2L, 2L, 2L),
c = c(11L, 13L, 13L, 16L, 15L, 156L)
)
df
#> year id c
#> 1 2012 1 11
#> 2 2013 1 13
#> 3 2014 1 13
#> 4 2012 2 16
#> 5 2013 2 15
#> 6 2014 2 156
widyr::pairwise_cor(df, id, year, c)
#> # A tibble: 2 x 3
#> item1 item2 correlation
#> <int> <int> <dbl>
#> 1 2 1 0.4946525
#> 2 1 2 0.4946525
widyr::pairwise_cor(df, id, year, c, upper = FALSE)
#> # A tibble: 1 x 3
#> item1 item2 correlation
#> <int> <int> <dbl>
#> 1 1 2 0.4946525
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.