英文:
Applying Bayesian Changepoint Detection algorithm using bcp to a grouped data frame in R
问题
我有一个分组的数据框，我想应用bcp函数来计算每个点的变化后验概率。
我的数据如下：
# 安装PACMAN
if (!require("pacman", character.only = TRUE)) {
install.packages("pacman")
}
pacman::p_load(bcp,tidyverse)
df <- data.frame(
date = c(seq(Sys.Date(), by = -1, length.out = 1000), seq(Sys.Date(), by = -1, length.out = 1000)),
value = c(rnorm(200, mean = 20, sd = 1), rnorm(800, mean = 17, sd = 2), rnorm(400, mean = 200, sd = 3), rnorm(600, mean = 150, sd = 4)),
product = c(rep("A", 1000), rep("B", 1000))
)
通过将我的数据框筛选为单个变量，并将其分配给新变量，然后应用bcp()函数，我看到它返回一个包含12个列表的结果：
x <- df %>%
filter(product == "A")
y <- bcp(x$value)
我尝试使用group_map，它返回了只有两列的结果，这可能不是理想的。不过，我不知道为什么只返回两列：
df %>%
group_by(product) %>%
group_map(~ bcp(.x$value))
我还尝试了group_modify，但无法得到正确的语法来解析出正确的字段：
df %>%
group_by(product) %>%
group_modify(~ {
bcp::bcp(.x$value) %>%
tibble::enframe(name = "name", value = "value")
})
以及：
df %>%
group_by(product) %>%
group_modify(~ bcp::bcp(.x$value) %>%
pluck("posterior.prob"))
关于如何在每个分组的基础上将bcp函数的'posterior.prob'附加到我的原始数据框上，如果有任何指导，将不胜感激。
英文:
I have a grouped dataframe which I would like to apply the bcp function to calculate for each point the posterior probability of there being a change at each point.
My data looks as follows:
# INSTALL PACMAN
if (!require("pacman", character.only = TRUE)) {
install.packages("pacman")
}
pacman::p_load(bcp,tidyverse)
df <- data.frame(
date = c(seq(Sys.Date(), by = -1, length.out = 1000), seq(Sys.Date(), by = -1, length.out = 1000)),
value = c(rnorm(200, mean = 20, sd = 1), rnorm(800, mean = 17, sd = 2), rnorm(400, mean = 200, sd = 3), rnorm(600, mean = 150, sd = 4)),
product = c(rep("A", 1000), rep("B", 1000))
)
By filtering my df to a single variable and assigning it to a new variable and applying bcp() I see it returns a list of 12
x <- df %>%
filter(product == "A")
y <- bcp(x$value)
I've tried using group_map which returns only two columns which is not ideal, I've no idea why only two columns are returned:
df %>%
group_by(product) %>%
group_map(~ bcp(.x$value))
I've also tried group_modify but I can't get the syntax correct to parse out the correct fields:
df %>%
group_by(product) %>%
group_modify(~ {
bcp::bcp(.x$value) %>%
tibble::enframe(name = "name", value = "value")
})
As well as:
df %>%
group_by(product) %>%
group_modify(~ bcp::bcp(.x$value) %>%
pluck("posterior.prob"))
Any guidance on how I can append the 'posterior.prob' from the bcp function to my original df on a per group basis would be greatly appreciated.
答案1
得分: 1
I'm not familiar with the bcp package but does this give you what you want?
posterior_prob <- map(df %>%
group_by(product) %>%
group_map(~ bcp(.x$value)), pluck("posterior.prob")) %>%
unlist()
df$posterior_prob_var <- posterior_prob
head(df)
# date value product posterior_prob_var
# 1 2023-05-10 21.90542 A 0.002
# 2 2023-05-09 19.61293 A 0.000
# 3 2023-05-08 20.46336 A 0.002
# 4 2023-05-07 21.22534 A 0.000
# 5 2023-05-06 19.37578 A 0.000
# 6 2023-05-05 18.94408 A 0.002
(Note: This is the translated code section. If you have any further questions or need additional assistance, please let me know.)
英文:
I'm not familiar with the bcp package but does this give you what you want?
posterior_prob <- map(df %>%
group_by(product) %>%
group_map(~ bcp(.x$value)), pluck("posterior.prob")) %>%
unlist()
df$posterior_prob_var <- posterior_prob
head(df)
# date value product posterior_prob_var
# 1 2023-05-10 21.90542 A 0.002
# 2 2023-05-09 19.61293 A 0.000
# 3 2023-05-08 20.46336 A 0.002
# 4 2023-05-07 21.22534 A 0.000
# 5 2023-05-06 19.37578 A 0.000
# 6 2023-05-05 18.94408 A 0.002
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论