Confidence Intervals in R by splitting column-CodePudding

How would you split a column based on name and take the confidence Intervals split by specific names?

Can't Index because importing a file a range, because importing different files. Can't hardcode names

Data Looks like:

Name | Score

Anna   90
Anna   90
Anna   30
Anna   60
Anna   60
Anna   60
Anna   60 
Bob    80
Bob    70
Bob    10
Bob    80 
Chad   10  
Chad   10 
Chad   40 
Chad   30 
Chad   90

How would you take the confidence intervals for

Anna | Bob | Chad

Tried splitting

#df[c('Name')] <- str_split_fixed(df, ' ', 1)

Tried tapply

CodePudding user response：

You were close with tapply. Either look in the textbook and code a ci function like this,

ci <- \(x, a=.05) {
  p <- sapply(list(length, sd, mean), \(f) f(x))
  tt <- -qt(a/2, p[1] - 1)
  p[3]   tt*(p[2]/sqrt(p[1]))*c(-1, 1)
}

with(dat, tapply(Score, Name, ci)) |> do.call(what=rbind)
#           [,1]      [,2]
# Anna 45.139588  83.43184
# Bob   6.431446 113.56855
# Chad -4.805243  76.80524

or extract the CIs from t.tests.

with(dat, tapply(Score, Name, t.test)) |> lapply(`[[`, 'conf.int') |> 
  rbind.data.frame() |> t()
#           [,1]      [,2]
# Anna 45.139588  83.43184
# Bob   6.431446 113.56855
# Chad -4.805243  76.80524

Data:

dat <- structure(list(Name = c("Anna", "Anna", "Anna", "Anna", "Anna", 
"Anna", "Anna", "Bob", "Bob", "Bob", "Bob", "Chad", "Chad", "Chad", 
"Chad", "Chad"), Score = c(90L, 90L, 30L, 60L, 60L, 60L, 60L, 
80L, 70L, 10L, 80L, 10L, 10L, 40L, 30L, 90L)), class = "data.frame", row.names = c(NA, 
-16L))

CodePudding user response：

Something like this based on split

sapply(split(df, df$Name), 
  function(x){ 
    mn    <- mean(x$Score)
    ln    <- length(x$Score)
    sdd   <- sd(x$Score)
    sd_er <- sdd / sqrt(ln)
    alpha <- 0.05
    dof   <- ln - 1
    t_score <- qt(p=alpha / 2, df=dof, lower.tail=F)
    mg_er <- t_score * sd_er
    l_bound <- mn - mg_er
    u_bound <- mn   mg_er
    c(lb=l_bound, ub=u_bound)})
       Anna        Bob      Chad
lb 45.13959   6.431446 -4.805243
ub 83.43184 113.568554 76.805243

With dplyr

library(dplyr)

df %>% 
  group_by(Name) %>% 
  summarize(mn = mean(Score), ln = length(Score), sdd = sd(Score),
    sd_er = sdd / sqrt(ln), alpha = 0.05, dof = ln - 1, 
    t_score = qt(p=alpha / 2, df=dof, lower.tail=F), 
    mg_er = t_score * sd_er) %>% 
  summarize(Name, l_bound = mn - mg_er, u_bound = mn   mg_er)
# A tibble: 3 × 3
  Name  l_bound u_bound
  <chr>   <dbl>   <dbl>
1 Anna    45.1     83.4
2 Bob      6.43   114.
3 Chad    -4.81    76.8

Data

df <- structure(list(Name = c("Anna", "Anna", "Anna", "Anna", "Anna",
"Anna", "Anna", "Bob", "Bob", "Bob", "Bob", "Chad", "Chad", "Chad",
"Chad", "Chad"), Score = c(90, 90, 30, 60, 60, 60, 60, 80, 70,
10, 80, 10, 10, 40, 30, 90)), class = "data.frame", row.names = c(NA,
-16L))