#set document(title: "6.5 The Central Limit Theorem", author: "Rachel Webb")
#set page(width: 8.5in, height: auto, margin: 1in)
#import "@preview/cetz:0.5.2"
#set text(font: ("STIX Two Text", "Libertinus Serif", "New Computer Modern"), size: 10.5pt, lang: "en")
#show math.equation: set text(font: ("STIX Two Math", "New Computer Modern Math"))
#set par(justify: true, leading: 0.62em, spacing: 0.9em)
#set enum(spacing: 1.1em)   // room between list items so tall inline fractions don't collide
#set list(spacing: 1.1em)
#set table(stroke: 0.5pt + rgb("#c7ccd3"))

#let BLUE = rgb("#183B6F")     // brand navy — section bars + example/solution labels (white on navy 11.09:1)
#let ORANGE = rgb("#A94509")   // brand primary-700 — AA-safe deep orange for TEXT (5.93:1 on white; raw brand #F37021 is 2.94:1 and must never carry text)
#let RED = rgb("#DC2626")      // brand error-600
#let GREEN = rgb("#059669")    // brand success-600 (decoration only; small green text uses green-text #007942)

#show heading.where(level: 1): it => block(width: 100%, above: 0pt, below: 16pt,
  fill: gradient.linear(BLUE, rgb("#2C5AA0")), inset: (x: 14pt, y: 12pt), radius: 3pt,
  text(fill: white, weight: "bold", size: 19pt, it.body))
#show heading.where(level: 2): it => block(width: 100%, above: 18pt, below: 10pt,
  fill: BLUE, inset: (x: 10pt, y: 6pt), radius: 2pt,
  text(fill: white, weight: "bold", size: 12pt, it.body))
#show heading.where(level: 3): it => text(fill: ORANGE, weight: "bold", size: 12.5pt, it.body)
#show heading.where(level: 4): it => text(fill: BLUE, weight: "bold", size: 10.5pt, it.body)

#let examplebox(label, title, body) = block(width: 100%, breakable: true,
  fill: rgb("#EFF1F5"), stroke: 0.5pt + rgb("#CFDDF0"), radius: 4pt, inset: 10pt,
  above: 12pt, below: 12pt)[
  #block(below: 6pt)[#box(fill: BLUE, inset: (x: 6pt, y: 2pt), radius: 2pt,
    text(fill: white, weight: "bold", size: 8.5pt, label)) #h(0.4em) #strong[#title]]
  #body]

// rail = decorative left rule (raw brand token); labelcolor = AA-safe label text shade
#let notebox(label, rail, labelcolor, tint, body) = block(width: 100%, breakable: true, fill: tint,
    stroke: (left: 3pt + rail), inset: (left: 10pt, rest: 8pt), radius: (right: 4pt),
    above: 11pt, below: 11pt)[
    #text(fill: labelcolor, weight: "bold", size: 7.5pt, tracking: 0.5pt)[#upper(label)]
    #linebreak() #body]

#let solutionbox(body) = block(above: 4pt, below: 8pt)[
  #text(fill: BLUE, weight: "bold", size: 8.5pt)[Solution] #linebreak() #body]

#let figph(msg) = block(width: 100%, height: 60pt, fill: rgb("#f6f7f9"),
  stroke: (paint: rgb("#c7ccd3"), dash: "dashed"), radius: 4pt, inset: 10pt)[
  #align(center + horizon, text(fill: rgb("#889"), style: "italic", size: 9pt, msg))]

// Standardize inlined figure sizes: measure the natural CeTZ canvas, then scale to a
// consistent envelope (aspect-aware; see build_typst.py FIG_* constants). Unlike the
// print preamble, dimensions are FLOORED: in an editor a user can trim a figure to a
// degenerate 1-D shape (a bare line), and w/h or tw/w would then divide by zero.
#let _STD_W = 3.5
#let _WIDE_W = 5.6
#let _MAX_H = 3.4
#let _ASPECT_WIDE = 2.2
#let _UPSCALE_MAX = 1.15
#let stdfig(body) = context {
  let m = measure(body)
  let w = calc.max(m.width / 1in, 0.01)
  let h = calc.max(m.height / 1in, 0.01)
  let tw = if w / h > _ASPECT_WIDE { _WIDE_W } else { _STD_W }
  let s = calc.min(tw / w, _MAX_H / h, _UPSCALE_MAX)
  align(center, box(scale(x: s * 100%, y: s * 100%, reflow: true, body)))
}

#show figure: set block(breakable: false)
#set figure(gap: 8pt)
#show figure.caption: set text(size: 8.5pt, fill: rgb("#555"))


    
== 6.5#h(0.6em)The Central Limit Theorem


    The sample mean, denoted #math.equation(block: false, alt: "x bar")[$limits(x)^(―)$], is the average of a sample of a variable X. The sample mean is an estimate of the population mean µ. Every sample has a sample mean and these sample means differ (depending on the sample). Thus, before a sample is selected #math.equation(block: false, alt: "x bar")[$limits(x)^(―)$] is a variable, in fact, if the sample is a random sample then #math.equation(block: false, alt: "x bar")[$limits(x)^(―)$] is a random variable. For this reason, we can think of the “distribution of #math.equation(block: false, alt: "x bar")[$limits(x)^(―)$],” called the “Sampling Distribution of #math.equation(block: false, alt: "x bar")[$limits(x)^(―)$],” as the theoretical histogram constructed from the sample averages of all possible samples of size n.Definition: Word


    #notebox("Note", rgb("#8a94a6"), rgb("#556666"), rgb("#f7f8fa"))[
Mean and Standard Deviation of a Sample Mean


      Let #math.equation(block: false, alt: "x bar")[$limits(x)^(―)$] be the mean of a random sample of size n from a population having mean μ and standard deviation σ, then


      The mean of the sample means = #math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] = µ.


      The standard deviation (standard error) of the sample means = #math.equation(block: false, alt: "σ sub x bar equals the fraction σ over the square root of n")[$σ_(overline(x)) = frac(σ, sqrt(n))$].
]


    This says that the mean of the sample means is the same as the population mean. The standard deviation of the sample means is the population standard deviation divided by the square root of the sample size. This is called the #strong[sampling distribution] of the mean.


    #examplebox("Example 1")[][
Let X be the height of men in the United States. Studies show that the heights of 15-year old boys in the United States are normally distributed with average height 67 inches and a standard deviation of 2.5 inches. A random experiment consists of choosing 16 15-year old boys at random. Compute the mean and standard deviation of #math.equation(block: false, alt: "x bar")[$limits(x)^(―)$], that is, the mean and standard deviation for the average height of a random sample of 16 boys.


        \# Simulate the sampling distribution of x-bar: N(67, 2.5), n = 16
set.seed(1)
xbar \<- replicate(10000, mean(rnorm(16, 67, 2.5)))
c(mean = mean(xbar), sd = sd(xbar))   \# ~67 and ~0.625
hist(xbar, main = "Sampling distribution of x-bar (n = 16)")
      
      #solutionbox[
The mean of the sample means is the same as the population mean #math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] = 67.


        The standard deviation in the sample means is the population standard deviation divided by the square root of the sample size, #math.equation(block: false, alt: "σ sub x bar equals the fraction σ over the square root of n equals the fraction 2.5 over the square root of 16 equals 0.625")[$σ_(overline(x)) = frac(σ, sqrt(n)) = frac(2.5, sqrt(16)) = 0.625$].


        Notice that the mean of a sample means is always the same as the mean of the population, but the standard deviation is smaller. See Figure 6-30.


        #figure(figph[Two normal curves centered at 67 on an axis from 62 to 72: a tall narrow red curve labeled N(67, 0.625) and a lower wider blue curve labeled N(67, 2.5).], alt: "Two normal curves centered at 67 on an axis from 62 to 72: a tall narrow red curve labeled N(67, 0.625) and a lower wider blue curve labeled N(67, 2.5).", caption: [Figure 6-30])
]
]


    #notebox("Note", rgb("#8a94a6"), rgb("#556666"), rgb("#f7f8fa"))[
#strong[Sampling Distribution of a Sample Mean]


      If a population is normally distributed N(µ, σ), then the sample mean #math.equation(block: false, alt: "x bar")[$limits(x)^(―)$] of n independent observations is normally distributed as #math.equation(block: false, alt: "N open parenthesis μ , the fraction σ over the square root of n close parenthesis")[$N ( μ , frac(σ, sqrt(n)) )$] /
]


    Figure 6-31 shows three population distributions and the corresponding sampling distributions for sample sizes of 2, 5, 12 and 30. Notice as the sample size gets larger, the sampling distribution gets closer to the dashed red line of the normal distribution. #strong[Video explanation of this process:] #link("https://youtu.be/lsCc_pS3O28.")[https://youtu.be/lsCc\_pS3O28.]


    #figure(figph[Grid of density plots for uniform, exponential, and log-normal populations: the top row shows each population distribution, and the rows below show sampling distributions of the mean for n = 2, 5, 12, and 30 in solid blue against a dashed red normal curve; by n = 30 all three closely match the normal curve.], alt: "Grid of density plots for uniform, exponential, and log-normal populations: the top row shows each population distribution, and the rows below show sampling distributions of the mean for n = 2, 5, 12, and 30 in solid blue against a dashed red normal curve; by n = 30 all three closely match the normal curve.", caption: none)


    Retrieved from #link("https://www.openintro.org/stat/textbook.php?stat_book=os")[OpenIntroStatistics].


    Figure 6-31


    The Central Limit Theorem establishes that in some situations the distribution of the sample statistic will take on a normal distribution, even when the population is not normally distributed. This allows us to use the normal distribution to make inferences from samples to populations.


    The Central Limit Theorem guarantees that the distribution of the sample mean will be normally distributed when the sample size is large (usually 30 or higher) no matter what shape the population distribution is.


    #strong[Finding Probabilities Using the Central Limit Theorem (CLT)]


    If we are finding the probability of a sample mean and have a sample size of 30 or more, or the population was normally distributed, then we can use the normal distribution to find the probability that the sample mean is below, above or between two values using the CLT.


    Watch this video on using this applet for the Central Limit Theorem, and then take some time to play with the applet to get a sense of the difference between the distribution of the population, the distribution of a sample and the sampling distribution.


    Watch the video on how to use the applet: #link("https://youtu.be/aIPvgiXyBMI")[https://youtu.be/aIPvgiXyBMI].


    Try the applet on your own. Applet: #link("http://onlinestatbook.com/stat_sim/sampling_dist/index.html")[http://onlinestatbook.com/stat\_sim/sampling\_dist/index.html].


    #examplebox("Example 2")[][
The population of midterm scores for all students taking a PSU Business Statistics course has a known standard deviation of 5.27. The mean of the population is 18.07 and the median of the population is 19. A sample of 25 was taken and the sample mean was 18.07 and we want to know what the sampling distribution for the mean looks like. Figure 6-32 shows 3 graphs using the Sampling Distribution Applet.


Figure 6-32


a) What is the mean and standard deviation of the sampling distribution?

b) Would you expect midterm exam scores to be skewed or bell-shaped?

c) Which of these graphs in Figure 6-32 correspond to the distribution of the population, distribution of a single sample and the sampling distribution of the mean?

d) Compute the probability that for next term’s class they have a sample mean of more than 20.

Solution

a) By the Central Limit Theorem (CLT) the mean of the sampling distribution #math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] equals the mean of the population which was given as µ=18.07. The standard deviation of the sampling distribution by the CLT would be the population standard deviation divided by the square root of the sample size #math.equation(block: false, alt: "σ sub x bar equals the fraction σ over the square root of n equals the fraction 5.27 over the square root of 25 equals 1.054")[$σ_(overline(x)) = frac(σ, sqrt(n)) = frac(5.27, sqrt(25)) = 1.054$]

b) The population mean = 18.07 is smaller than the median = 19 therefore the distribution is negatively skewed, the mean is pulled in the direction of the outliers.

c) Using the #link("http://onlinestatbook.com/stat_sim/sampling_dist/index.html")[Sampling Distribution Applet] and the CLT, the sampling distribution will be bell-shaped therefore, graph 3 has to be the sampling distribution.

Graphs 1 & 2 in Figure 6-32 are both negatively skewed. A single sample of 25 should look similar to the entire population, but we would expect only 25 items and not every score possible would be received from the 25 students.

Graph 1 in Figure 6-32 fits this description and therefore the graph of the distribution of a single sample (which is not the same thing as the sampling distribution) is graph 1.

This leaves graph 2 as the distribution of the population. Figure 6-33 is a picture of the applet modeling the exam scores. Note the top picture is the population distribution, the second graph is simulating a single sample drawn and the bottom picture is a graph of all the sample means for each sample. This last graph is the sampling distribution of the means.


Figure 6-33


d) The P(#math.equation(block: false, alt: "X bar")[$overline(X)$] \> 20) would be normally distributed with a mean #math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] = 18.07 with a standard deviation of #math.equation(block: false, alt: "σ sub x bar equals the fraction σ over the square root of n equals the fraction 5.27 over the square root of 25 equals 1.054")[$σ_(overline(x)) = frac(σ, sqrt(n)) = frac(5.27, sqrt(25)) = 1.054$]

Draw and shade the sampling distribution curve.

This calculator can be used to draw and shade the sampling distribution: #link("http://homepage.divms.uiowa.edu/~mbognar/applets/normal.html")[http://homepage.divms.uiowa.edu/~mbognar/applets/normal.html], filling in the mean μ, standard deviation #math.equation(block: false, alt: "the fraction σ over the square root of n")[$frac(σ, sqrt(n))$] and x-value (in this case the sample mean) will find the probability. See Figure 6-34.


Figure 6-34


TI Calculator: normalcdf(20,1E99,18.07,5.27/√25) = 0.0335.

Excel: P(#math.equation(block: false, alt: "X bar")[$overline(X)$] \> 20) =1-NORM.DIST(20,18.07,5.27/SQRT(25),TRUE) = 0.0335.
]


    #examplebox("Example 3")[][
Let X be the height of 15-year old boys in the United States. Studies show that the heights of 15-year old boys in the United States are normally distributed with average height of 67 inches and a standard deviation of 2.5 inches. A random experiment consists of randomly choosing sixteen 15-year old boys. Compute the probability that the mean height of those sampled is 69.5 inches or taller.


      
      #solutionbox[
The sample mean #math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] is approximately #math.equation(block: false, alt: "N open parenthesis 67 , 0.625 close parenthesis . P open parenthesis X bar greater than or equal to 69.5 close parenthesis equals P open parenthesis the fraction 8 minus 67 over 0.625 greater than or equal to the fraction 69.5 minus 67 over 0.625 close parenthesis equals P open parenthesis Z greater than or equal to 4 close parenthesis approximately equals 0.00003")[$N ( 67 , 0.625 ) . P ( overline(X) ≥ 69.5 ) = P ( frac(8 − 67, 0.625) ≥ frac(69.5 − 67, 0.625) ) = P ( Z ≥ 4 ) ≈ 0.00003$], using the calculator, be careful with the scientific notation. This is a very small probability.


        #figure(figph[TI-84 screen showing normalcdf(69.5,1E99,67,2.5/sqrt(16)) with the result 3.168603459E-5.], alt: "TI-84 screen showing normalcdf(69.5,1E99,67,2.5/sqrt(16)) with the result 3.168603459E-5.", caption: none)


        This should make sense because one would think that the likelihood of randomly selecting 16 boys that have an average height of 5’9.5” would be slim.


        Figure 6-35 shows the density curves showing the shaded areas of P(X ≥ 69.5) and P(#math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] ≥ 69.5).


        #figure(figph[Two normal curves centered at 67 on an axis from 62 to 72: the wide blue N(67, 2.5) curve has its tail beyond 69.5 shaded and labeled Area = 0.1587, while the tall narrow red N(67, 0.625) curve has a barely visible tail labeled Area = 0.00003 with an arrow at 69.5.], alt: "Two normal curves centered at 67 on an axis from 62 to 72: the wide blue N(67, 2.5) curve has its tail beyond 69.5 shaded and labeled Area = 0.1587, while the tall narrow red N(67, 0.625) curve has a barely visible tail labeled Area = 0.00003 with an arrow at 69.5.", caption: [Figure 6-35])


        The sampling distribution has a much smaller spread (standard deviation) and hence less area to the right of 69.5.
]
]


    #notebox("Note", rgb("#8a94a6"), rgb("#556666"), rgb("#f7f8fa"))[
In general, the Central Limit Theorem questions will use the same method as previous sections, however you will use a standard deviation of #math.equation(block: false, alt: "the fraction σ over the square root of n")[$frac(σ, sqrt(n))$] and a z-score of #math.equation(block: false, alt: "z equals the fraction x bar minus μ over open parenthesis the fraction σ over the square root of n close parenthesis")[$z = frac(overline(x) − μ, ( frac(σ, sqrt(n)) ))$].
]


    #examplebox("Example 4")[][
The average teacher’s salary in Connecticut (ranked first among states) is \$57,337. Suppose that the distribution of salaries is normally distributed with a standard deviation of \$7,500.


        mu \<- 57337; sigma \<- 7500
pnorm(55000, mu, sigma)            \# one teacher   -\> 0.3777
pnorm(55000, mu, sigma/sqrt(10))   \# mean of n=10  -\> 0.1622
pnorm(55000, mu, sigma/sqrt(100))  \# mean of n=100 -\> 0.00092
        #notebox("Note", rgb("#8a94a6"), rgb("#556666"), rgb("#f7f8fa"))[
#emph[Watch the probability fall as n grows]


          Pre-loaded with one teacher (sd 7500): P(X \< 55000) = 0.3777. Re-enter the standard deviation as 7500/sqrt(10) = 2371.71, then 7500/sqrt(100) = 750, and watch P fall to 0.1622 and then 0.00092.


          - Teacher salaries N(57337, 7500): P(X \< 55000) = 0.3777
]


        a) What is the probability that a randomly selected teacher makes less than \$55,000 per year?


        b) If we sample 10 teachers’ salaries, what is the probability that the sample mean is less than \$55,000?


        c) If we sample 100 teachers’ salaries, what is the probability that the sample mean is less than \$55,000?


      
      #solutionbox[
a) Find P(X \< 55000), since we are only looking at one person use #math.equation(block: false, alt: "z equals the fraction x minus μ over σ")[$z = frac(x − μ, σ)$]. If we were asked to standardize the salary, #math.equation(block: false, alt: "z equals the fraction 55000 minus 57337 over 7500 equals minus 0.3116")[$z = frac(55000 − 57337, 7500) = − 0.3116$], however we can use technology and skip this step.


        Use the normalcdf(-1E99,55000,57337,7500) (the TI-89 use -∞ for the lower boundary instead of -1E99) and you get the probability of 0.3777. The P(X \< 55000) = P(Z \< –0.3116) = 0.3777.


        Note that we are not using the CLT since we are not finding the probability of an average for a group of people, just the probability for one person.


        #figure(figph[TI-84 screen showing normalcdf(-1E99,55000,57337,7500) with the result .3776723318.], alt: "TI-84 screen showing normalcdf(-1E99,55000,57337,7500) with the result .3776723318.", caption: none)


        b) Find P(#math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] \< 55000), but we are looking at the probability of a mean for 10 teachers, so use #math.equation(block: false, alt: "z equals the fraction x bar minus μ over open parenthesis the fraction σ over the square root of n close parenthesis")[$z = frac(overline(x) − μ, ( frac(σ, sqrt(n)) ))$]. Standardize the salary, #math.equation(block: false, alt: "z equals the fraction 55000 minus 57337 over open parenthesis the fraction 7500 over the square root of 10 close parenthesis equals – 0.9853657189")[$z = frac(55000 − 57337, ( frac(7500, sqrt(10)) )) = – 0.9853657189$], use your calculator to get P(#math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] \< 55000) = P(Z \< –0.9853657189) = 0.1622.


        You do not need the extra step of finding the z-score first. Instead you can use normalcdf(-1E99,55000,57337,7500/√10) = 0.1622.


        #figure(figph[TI-84 screen showing (55000-57337)/(7500/sqrt(10)) = -.9853657189, followed by normalcdf(-1E99,-.9853657189,0,1) returning .162222231.], alt: "TI-84 screen showing (55000-57337)/(7500/sqrt(10)) = -.9853657189, followed by normalcdf(-1E99,-.9853657189,0,1) returning .162222231.", caption: none)


        c) Find P(#math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] \< 55000), but we are looking at the probability of a mean for 100 teachers, so use #math.equation(block: false, alt: "z equals the fraction x bar minus μ over open parenthesis the fraction σ over the square root of n close parenthesis")[$z = frac(overline(x) − μ, ( frac(σ, sqrt(n)) ))$]. Standardize the salary, #math.equation(block: false, alt: "z equals the fraction 55000 minus 57337 over open parenthesis the fraction 7500 over the square root of 100 close parenthesis equals – 3.116")[$z = frac(55000 − 57337, ( frac(7500, sqrt(100)) )) = – 3.116$] use your calculator normalcdf(-1E99,55000,57337,7500/√100) to get P(#math.equation(block: false, alt: "μ sub x bar")[$μ_(overline(x))$] \< 55000) = P(Z \< –3.116) = 0.0009167.


        #figure(figph[TI-84 screen showing (55000-57337)/(7500/sqrt(100)) = -3.116, followed by normalcdf(-1E99,-3.116,0,1) returning 9.166789234E-4.], alt: "TI-84 screen showing (55000-57337)/(7500/sqrt(100)) = -3.116, followed by normalcdf(-1E99,-3.116,0,1) returning 9.166789234E-4.", caption: none)
]
]


    As the sample size increase, the probability of seeing a sample mean of less than \$55,000 is getting smaller.


    When you have a z-score that is less than –3 or greater than 3 we would call this a rare event or outlier. We will use this same process in inferential statistics in chapter 8.