Skip to main content
Statistics LibreTexts

13.2: Central Limit Theorem

  • Page ID
    8791
  • \( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \) \( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)\(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\) \(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\)\(\newcommand{\AA}{\unicode[.8,0]{x212B}}\)

    The central limit theorem tells us that the sampling distribution of the mean becomes normal as the sample size grows. Let’s test this by sampling a clearly non-normal variable and look at the normality of the results using a Q-Q plot. We saw in Figure @ref{fig:alcDist50} that the variable AlcoholYear is distributed in a very non-normal way. Let’s first look at the Q-Q plot for these data, to see what it looks like. We will use the stat_qq() function from ggplot2 to create the plot for us.

    # prepare the dta
    NHANES_cleanAlc <- NHANES %>%
      drop_na(AlcoholYear)
    
    ggplot(NHANES_cleanAlc, aes(sample=AlcoholYear)) +
      stat_qq() + 
      # add the line for x=y
      stat_qq_line()

    file70.png

    We can see from this figure that the distribution is highly non-normal, as the Q-Q plot diverges substantially from the unit line.

    Now let’s repeatedly sample and compute the mean, and look at the resulting Q-Q plot. We will take samples of various sizes to see the effect of sample size. We will use a function from the dplyr package called do(), which can run a large number of analyses at once.

    set.seed(12345)
    
    sampSizes <- c(16, 32, 64, 128) # size of sample
    nsamps <- 1000 # number of samples we will take
    
    # create the data frame that specifies the analyses
    input_df <- tibble(sampSize=rep(sampSizes,nsamps),
                          id=seq(nsamps*length(sampSizes)))
    
    
    # create a function that samples and returns the mean
    # so that we can loop over it using replicate()
    get_sample_mean <- function(sampSize){
      meanAlcYear <- 
        NHANES_cleanAlc %>%
        sample_n(sampSize) %>%
        summarize(meanAlcoholYear = mean(AlcoholYear)) %>%
        pull(meanAlcoholYear)
      return(tibble(meanAlcYear = meanAlcYear, sampSize=sampSize))
    }
    
    # loop through sample sizes
    # we group by id so that each id will be run separately by do()
    all_results = input_df %>% 
      group_by(id) %>%
      # "." refers to the data frame being passed in by do()
      do(get_sample_mean(.$sampSize))

    Now let’s create separate Q-Q plots for the different sample sizes.

    # create empty list to store plots
    
    qqplots = list()
    
    for (N in sampSizes){
      sample_results <- 
        all_results %>%
        filter(sampSize==N)
    
      qqplots[[toString(N)]] <- ggplot(sample_results, 
                                    aes(sample=meanAlcYear)) +
        stat_qq() + 
        # add the line for x=y
        stat_qq_line(fullrange = TRUE) + 
        ggtitle(sprintf('N = %d', N)) + 
       xlim(-4, 4) 
    
    }
    
    plot_grid(plotlist = qqplots)

    file71.png

    This shows that the results become more normally distributed (i.e. following the straight line) as the samples get larger.


    This page titled 13.2: Central Limit Theorem is shared under a not declared license and was authored, remixed, and/or curated by Russell A. Poldrack via source content that was edited to the style and standards of the LibreTexts platform; a detailed edit history is available upon request.