通过数字变量过滤数据帧lm（）并提取斜率

我想应用不同级别的数值过滤器（例如seq(10,80, by=2)），然后将它们拼接回单个数据框以与另一个变量进行比较。我目前可以做到这一点，但我希望有一种更好的方法，因为我只是复制并粘贴代码，然后再加入所有内容。我想要的最终结果是我所拥有的，每个过滤步骤都作为其自己的列，并且从lm（）中提取斜率参数。通过数字变量过滤数据帧lm（）并提取斜率

Source: local data frame [23 x 17] 

          File FruitNum  est10 
         <fctr> <int>  <dbl> 
1 IMG_7888.JPGcolcorrected.jpg  2 -4.0000000 
2 IMG_7888.JPGcolcorrected.jpg  4 -2.0000000 
3 IMG_7889.JPGcolcorrected.jpg  1 -0.8178571 
4 IMG_7889.JPGcolcorrected.jpg  2 -2.1000000 
5 IMG_7890.JPGcolcorrected.jpg  1 -2.8000000 
6 IMG_7892.JPGcolcorrected.jpg  3 -2.3571429 
7 IMG_7895.JPGcolcorrected.jpg  1 -0.4000000 
8 IMG_7896.JPGcolcorrected.jpg  3 -6.5000000 
9 IMG_7898.JPGcolcorrected.jpg  1 -3.0000000 
10 IMG_7888.JPGcolcorrected.jpg  1   NA 
..       ...  ...  ... 
Variables not shown: est15 <dbl>, est20 <dbl>, est25 <dbl>, 
    est30 <dbl>, est35 <dbl>, est40 <dbl>, est45 <dbl>, est50 
    <dbl>, est55 <dbl>, est60 <dbl>, est65 <dbl>, est70 <dbl>, 
    est75 <dbl>, est80 <dbl>.

我目前使用的hadleyverse的NSE管道，想呆在那里，但我高兴地看到基地，data.table或其他实现。我一直在看purrr，但我不确定如何将过滤器映射到内联变量。

library(dplyr) 
library(purrr) 
library(tidyr) 
library(broom) 

cukeDataDL <- read.delim("https://gist.githubusercontent.com/bhive01/e7508f552db0415fec1749d0a390c8e5/raw/a12386d43c936c2f73d550dfdaecb8e453d19cfe/widthtest.tsv") 

cukeDatatest <- 
    cukeDataDL %>% 
    mutate(ObjectWidth = strsplit(as.character(cukeDatatest$ObjectWidth), ',')) %>% # split ObjectWidth into a nested column containing a vector 
    unnest() %>% # unnest nested column, melting data to long form 
    mutate(ObjectWidth = as.integer(ObjectWidth)) %>% # convert data to integer 
    group_by(File, FruitNum) %>% 
    mutate(rownum = row_number()) #location within File x fruit 

estimate10 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.10 * max(ObjectWidth) & rownum > mean(rownum)) %>% # filtering for 10% of maxwidth and second half of fruit 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% #broom to clean up models and get coef()s 
    unnest() %>% #pull out nested information 
    filter(term == "rownum") %>% #only interested in slope value 
    select(File, FruitNum, est10 = estimate) #get rid of uninteresting columns and rename estimate for join 

estimate15 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.15 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est15 = estimate) 

estimate20 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.20 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est20 = estimate) 

estimate25 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.25 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est25 = estimate) 

estimate30 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.30 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est30 = estimate) 

estimate35 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.35 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est35 = estimate) 

estimate40 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.40 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est40 = estimate) 

estimate45 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.45 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est45 = estimate) 

estimate50 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.50 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est50 = estimate) 

estimate55 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.55 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est55 = estimate) 

estimate60 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.60 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est60 = estimate) 

estimate65 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.65 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est65 = estimate) 

estimate70 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.70 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est70 = estimate) 

estimate75 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.75 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est75 = estimate) 
estimate80 <- 
    cukeDatatest %>% 
    filter(ObjectWidth < 0.80 * max(ObjectWidth) & rownum > mean(rownum)) %>% 
    by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
    unnest() %>% 
    filter(term == "rownum") %>% 
    select(File, FruitNum, est80 = estimate) 

    # put everything together 
allEstimates <- 
    full_join(estimate10, estimate15) %>% 
    full_join(., estimate20) %>% 
    full_join(., estimate25) %>% 
    full_join(., estimate30) %>% 
    full_join(., estimate35) %>% 
    full_join(., estimate40) %>% 
    full_join(., estimate45) %>% 
    full_join(., estimate50) %>% 
    full_join(., estimate55) %>% 
    full_join(., estimate60) %>% 
    full_join(., estimate65) %>% 
    full_join(., estimate70) %>% 
    full_join(., estimate75) %>% 
    full_join(., estimate80) 
allEstimates #print it out

来源

2016-05-17 bhive01

这将是更好，如果你是在什么你试图做而不是显示你是怎么做到的更加清晰。为样本输入提供所需的输出。 – MrFlick

感谢您的评论@MrFlick。输出是所需的输出。我想要的帮助是从我的代码中删除所有重复。我确信它可以完成，我只是不知道从哪里开始。为了清晰起见，我重构了代码以使其更短，并对描述进行了编辑。 – bhive01

更短！感谢@NoamRoss通过twitter。

使用的地图，你提供你想要遍历seq(10,80, by=2)
它为每个迭代
一系列dataframes的创建namesafe列描述使用列名后
使用bind_rows系列（）将所有东西放在一起
使用spread（）使PCTwidth的每个级别成为一列
Profit ???

library(dplyr) 
library(purrr) 
library(tidyr) 
library(broom) 

cukeDataDL <- read.delim("https://gist.githubusercontent.com/bhive01/e7508f552db0415fec1749d0a390c8e5/raw/a12386d43c936c2f73d550dfdaecb8e453d19cfe/widthtest.tsv") 
cukeDatatest <- 
    cukeDataDL %>% 
     select(File, FruitNum, ObjectWidth) %>% 
     # split ObjectWidth into a nested column containing a vector 
     mutate(ObjectWidth = strsplit(as.character(.$ObjectWidth), ',')) %>% 
     # unnest nested column, melting data to long form 
     unnest() %>% 
     # convert data to integer 
     mutate(ObjectWidth = as.integer(ObjectWidth)) %>% # convert data to integer 
     group_by(File, FruitNum) %>% 
     mutate(rownum = row_number()) 
allEstimates <- 
    map(seq(0.10,0.80, by=0.02), function(x) { 
     cukeDatatest %>% 
      filter(ObjectWidth < x * max(ObjectWidth) & rownum > mean(rownum)) %>% 
      by_slice(~tidy(lm(ObjectWidth ~ rownum, data = .))) %>% 
      unnest() %>% 
      filter(term == "rownum") %>% 
      select(File, FruitNum, estimate) %>% 
      mutate(PCTwidth = paste("est", round(x*100), sep="")) 
     } 
    ) %>% 
    bind_rows() %>% 
    spread(., PCTwidth, estimate) 

allEstimates #print everything out

来源

2016-05-17 15:03:57 bhive01

通过数字变量过滤数据帧lm（）并提取斜率

回答

相关问题