Reframing output of confidence intervals to combine mean, upper and lower values into one cell

huangapple
go评论88阅读模式
英文:

Reframing output of confidence intervals to combine mean, upper and lower values into one cell

问题

I use the code below to calculate the mean, upper and lower confidence intervals of multiple variables at once.

  1. library(gmodels)
  2. library(purrr)
  3. dfci <- df %>%
  4.   group_by(group) %>%
  5.   dplyr::summarize(across(everything(),
  6.   .fns = list(mean = ~ mean(.x, na.rm = TRUE, trim = 4),
  7.   ci = ~ ci(.x, confidence = 0.95, alpha = 0.05, na.rm = T))))
  8. #dfci <- dfci[-(13:16),] # remove additional rows
  9. write.csv(dfci, file="dfci.csv")

Sample data:

  1. Group| A_pre  |    A_post |  B_pre |  B_post 
  3. 0       20          21        20        23
  4. 1       30          10        19        11
  5. 2       10          53        30        34
  6. 1       22          32        25        20
  7. 2       34          40        32        30
  8. 0       30          50        NA        40
  9. 0       39          40        19        20
  10. 1       40          NA        20        20
  11. 2       50          10        20        10
  12. 0       34          23        30        10

I tried pivoting into long after the ci calculations but doesn't work:

  1. library(reshape2)
  3. dfci <- df %>%
  4.   group_by(group) %>%
  5.   summarize(across(everything(),
  6.                    .fns = list(mean = ~ mean(.x, na.rm = TRUE, trim = 4),
  7.                                ci = ~ ci(.x, confidence = 0.95, alpha = 0.05, na.rm = TRUE))))
  9. dfci <- melt(dfci, id.vars = "group")
  10. dfci <- dcast(dfci, group + variable ~ variable)
  12. write.csv(dfci, file = "dfi.csv", row.names = FALSE)
英文:

I use the code below to calculate the mean, upper and lower confidence intervals of multiple variables at once.

  1. library(gmodels)
  2. library(purrr)
  3. dfci &lt;- df %&gt;% 
  4.   group_by(group) %&gt;% 
  5.   dplyr::summarize(across(everything(),
  6.   .fns = list(mean = ~ mean(.x, na.rm = TRUE, trim = 4),
  7.   ci = ~ ci(.x, confidence = 0.95, alpha = 0.05, na.rm = T))))
  8. #dfci &lt;- dfci[-(13:16),] # remove additional rows
  9. write.csv(dfci, file=&quot;dfci.csv&quot;)

Sample data:

  1. Group| A_pre  |    A_post |  B_pre |  B_post 
  3. 0       20          21        20        23
  4. 1       30          10        19        11
  5. 2       10          53        30        34
  6. 1       22          32        25        20
  7. 2       34          40        32        30
  8. 0       30          50        NA        40
  9. 0       39          40        19        20
  10. 1       40          NA        20        20
  11. 2       50          10        20        10
  12. 0       34          23        30        10

As I have over 50 "pre" and "post" variables i.e., >100 variables, is it possible to combine the outputs from the three desired cells (mean, lower and upper ci) into one so I am not manually combining all of them?

I tried pivoting into long after the ci calculations but doesn't work:

  2. library(reshape2)
  4. dfci &lt;- df %&gt;%
  5.   group_by(group) %&gt;%
  6.   summarize(across(everything(),
  7.                    .fns = list(mean = ~ mean(.x, na.rm = TRUE, trim = 4),
  8.                                ci = ~ ci(.x, confidence = 0.95, alpha = 0.05, na.rm = TRUE))))
  10. dfci &lt;- melt(dfci, id.vars = &quot;group&quot;)
  11. dfci &lt;- dcast(dfci, group + variable ~ variable)
  13. write.csv(dfci, file = &quot;dfi.csv&quot;, row.names = FALSE)

答案1

得分: 2

以下是您提供的内容的翻译：

更新后的澄清：

我们可以使用自定义的ci函数来使用reframe：

  1. library(dplyr) #&gt;= dplyr 1.1.0
  2. df %&gt;%  
  3.   reframe(across(everything(), .fns = list(
  4.     mean = ~ mean(.x, na.rm = TRUE, trim = 4),
  5.     ci = ~ {
  6.       se &lt;- sqrt(var(.x, na.rm = TRUE) / sum(!is.na(.x)))
  7.       mean_val &lt;- mean(.x, na.rm = TRUE)
  8.       lower &lt;- mean_val - qt(0.975, df = sum(!is.na(.x))) * se
  9.       upper &lt;- mean_val + qt(0.975, df = sum(!is.na(.x))) * se
  10.      # c(lower, upper)
  11.       paste0(&quot;[&quot;, round(lower, 2), &quot;, &quot;, round(upper, 2), &quot;]&quot;)
  12.     }
  13.   )), .by = Group)
  1.   Group A_pre_mean       A_pre_ci A_post_mean       A_post_ci B_pre_mean       B_pre_ci B_post_mean     B_post_ci
  2. 1     0         32 [19.56, 41.94]        31.5  [14.18, 52.82]         20 [11.82, 34.18]        21.5 [5.93, 40.57]
  3. 2     1         30  [14.1, 47.24]        21.0 [-26.33, 68.33]         20 [15.43, 27.24]        20.0 [7.45, 26.55]
  4. 3     2         34 [-5.66, 68.33]        40.0  [-6.19, 74.85]         30 [15.52, 39.15]        30.0 [1.04, 48.29]
英文:

Update after clarification:

We can use reframe with a custom ci function:

  1. library(dplyr) #&gt;= dplyr 1.1.0
  2. df %&gt;%  
  3.   reframe(across(everything(), .fns = list(
  4.     mean = ~ mean(.x, na.rm = TRUE, trim = 4),
  5.     ci = ~ {
  6.       se &lt;- sqrt(var(.x, na.rm = TRUE) / sum(!is.na(.x)))
  7.       mean_val &lt;- mean(.x, na.rm = TRUE)
  8.       lower &lt;- mean_val - qt(0.975, df = sum(!is.na(.x))) * se
  9.       upper &lt;- mean_val + qt(0.975, df = sum(!is.na(.x))) * se
  10.      # c(lower, upper)
  11.       paste0(&quot;[&quot;, round(lower, 2), &quot;, &quot;, round(upper, 2), &quot;]&quot;)
  12.     }
  13.   )), .by = Group)
  1.   Group A_pre_mean       A_pre_ci A_post_mean       A_post_ci B_pre_mean       B_pre_ci B_post_mean     B_post_ci
  2. 1     0         32 [19.56, 41.94]        31.5  [14.18, 52.82]         20 [11.82, 34.18]        21.5 [5.93, 40.57]
  3. 2     1         30  [14.1, 47.24]        21.0 [-26.33, 68.33]         20 [15.43, 27.24]        20.0 [7.45, 26.55]
  4. 3     2         34 [-5.66, 68.33]        40.0  [-6.19, 74.85]         30 [15.52, 39.15]        30.0 [1.04, 48.29]

答案2

得分: 0

This code does the job:

  1. library(dplyr)
  3. dfci <- df %>%
  4.   group_by(group) %>%
  5.   summarise(across(everything(), list(
  6.     mean = ~ mean(., na.rm = TRUE, trim = 4),
  7.     ci = ~ { # 自定义的 CI 函数 
  8.       se <- sqrt(var(., na.rm = TRUE) / sum(!is.na(.)))
  9.       mean_val <- mean(., na.rm = TRUE)
  10.       lower <- mean_val - qt(0.975, df = sum(!is.na(.))) * se
  11.       upper <- mean_val + qt(0.975, df = sum(!is.na(.))) * se
  12.       paste0("[", round(lower, 2), ", ", round(upper, 2), "]")
  13.     }
  14.   ), .names = "{.col}_{.fn}")) %>%
  15.   ungroup()
英文:

Unfortunately the earlier answers did not work as they repeated the same ci throughout.

This code does the job:

  1. library(dplyr)
  3. dfci &lt;- df %&gt;%
  4.   group_by(group) %&gt;%
  5.   summarise(across(everything(), list(
  6.     mean = ~ mean(., na.rm = TRUE, trim = 4),
  7.     ci = ~ { # OWN CI FUNCTION 
  8.       se &lt;- sqrt(var(., na.rm = TRUE) / sum(!is.na(.)))
  9.       mean_val &lt;- mean(., na.rm = TRUE)
  10.       lower &lt;- mean_val - qt(0.975, df = sum(!is.na(.))) * se
  11.       upper &lt;- mean_val + qt(0.975, df = sum(!is.na(.))) * se
  12.       paste0(&quot;[&quot;, round(lower, 2), &quot;, &quot;, round(upper, 2), &quot;]&quot;)
  13.     }
  14.   ), .names = &quot;{.col}_{.fn}&quot;)) %&gt;%
  15.   ungroup()

huangapple
  • 本文由 发表于 2023年5月18日 01:17:16
  • 转载请务必保留本文链接：https://go.coder-hub.com/76274650.html
  • confidence-interval
  • pivot
  • r
  • reshape2
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定