Notes: Variance, Standard Deviation, CV, and Grouped Data

  • Variance and standard deviation are measures of dispersion in a data set.
  • There are two common contexts:
    • Population: parameters that describe the entire population.
    • Sample: statistics computed from a sample drawn from the population.
  • Variance and standard deviation definitions
    • Population variance and standard deviation
    • Population variance: ext{Var}( ext{population}) = rac{1}{n}\,
      \sum{i=1}^{n} (xi - \mu)^2 = rac{n\sum xi^2 - (\sum xi)^2}{n^2}
    • Population standard deviation: σ=Var(extpopulation)=n∑x<em>i2−(∑x</em>i)2n2\sigma = \sqrt{\text{Var}( ext{population})} = \sqrt{\frac{n\sum x<em>i^2 - (\sum x</em>i)^2}{n^2}}
    • Sample variance and standard deviation
    • Sample variance (unbiased estimator): s^2 = \frac{1}{n-1}\n\sum{i=1}^{n} (xi - \bar{x})^2 = \frac{n\sum xi^2 - (\sum xi)^2}{n(n-1)}
    • Sample standard deviation: s=s2=n∑x<em>i2−(∑x</em>i)2n(n−1)s = \sqrt{s^2} = \sqrt{\frac{n\sum x<em>i^2 - (\sum x</em>i)^2}{n(n-1)}}
  • Why the difference (degrees of freedom)
    • When estimating the mean \bar{x} from the same data, we lose one degree of freedom, hence the division by (n-1) for an unbiased estimate of the variance.
  • Shortcut (computing) formulas for standard deviation
    • For a sample (shortcut formula):
    • s=n∑x<em>i2−(∑x</em>i)2n(n−1)s = \sqrt{ \frac{n\sum x<em>i^2 - (\sum x</em>i)^2}{n(n-1)} }
    • For a population (shortcut formula):
    • σ=n∑x<em>i2−(∑x</em>i)2n2\sigma = \sqrt{ \frac{n\sum x<em>i^2 - (\sum x</em>i)^2}{n^2} }
  • Why use the shortcut formulas?
    • They reduce the amount of arithmetic when you calculate by hand.
    • Both formulas involve a sum of squares and a sum; subtraction occurs in both, which is convenient for mental math.
    • The shortcut formulas are algebraically equivalent to the definitions above.
  • How to compute by hand (procedural steps)
    • Data handling
    • Write data vertically (one column of x-values).
    • Compute:
      • The sum of x: ∑x<em>i=S</em>x\sum x<em>i = S</em>x
      • The sum of squares: ∑x<em>i2=S</em>x2\sum x<em>i^2 = S</em>{x^2}
      • The sample size: nn
      • The mean: xˉ=Sxn\bar{x} = \frac{S_x}{n}
    • Plug into the shortcut formulas to obtain variance and standard deviation:
    • Sample variance: s2=n S<em>x2−(S</em>x)2n(n−1)s^2 = \frac{n\,S<em>{x^2} - (S</em>x)^2}{n(n-1)}
    • Sample standard deviation: s=s2s = \sqrt{s^2}
    • Population variance: σ2=n S<em>x2−(S</em>x)2n2\sigma^2 = \frac{n\,S<em>{x^2} - (S</em>x)^2}{n^2}
    • Population standard deviation: σ=σ2\sigma = \sqrt{\sigma^2}
    • Practical tip: do not rely on only a single number; use a calculator or computer tool to handle the sums accurately.
  • Worked illustrative example (consistent dataset)
    • Dataset: x = {2, 4, 4, 4, 6}
    • n = 5; Sx = 2 + 4 + 4 + 4 + 6 = 20; S{x^2} = 4 + 16 + 16 + 16 + 36 = 88
    • Mean: xˉ=Sxn=205=4\bar{x} = \frac{S_x}{n} = \frac{20}{5} = 4
    • Sample variance: s2=n S<em>x2−(S</em>x)2n(n−1)=5⋅88−2025⋅4=440−40020=4020=2s^2 = \frac{n\,S<em>{x^2} - (S</em>x)^2}{n(n-1)} = \frac{5\cdot 88 - 20^2}{5\cdot 4} = \frac{440 - 400}{20} = \frac{40}{20} = 2
    • Sample standard deviation: s=2≈1.4142s = \sqrt{2} \approx 1.4142
    • Population variance: σ2=n S<em>x2−(S</em>x)2n2=440−40025=4025=1.6\sigma^2 = \frac{n\,S<em>{x^2} - (S</em>x)^2}{n^2} = \frac{440 - 400}{25} = \frac{40}{25} = 1.6
    • Population standard deviation: σ=1.6≈1.2649\sigma = \sqrt{1.6} \approx 1.2649
    • Coefficient of variation (sample): CV=sxˉ=1.41424≈0.3536⇒35.36%CV = \frac{s}{\bar{x}} = \frac{1.4142}{4} \approx 0.3536 \Rightarrow 35.36\%
  • Coefficient of variation (CV)
    • Definition: a dimensionless measure of relative dispersion
    • For a sample: CV=sxˉCV = \frac{s}{\bar{x}}
    • For a population: CVpop=σμCV_{pop} = \frac{\sigma}{\mu} where μ=xˉ\mu = \bar{x} in the population context
    • Interpretation: Compare variability across data sets with different means or different units; a higher CV means greater relative variability
    • Example interpretation (based on the illustrative data): CV around 35.36% indicates a moderate level of dispersion relative to the mean
  • Relationship between mean and variation (intuition)
    • The mean describes central tendency; the standard deviation describes spread around the mean.
    • A data set with the same mean but a larger standard deviation shows more spread and less consistency.
    • Conversely, a smaller standard deviation indicates the data are more tightly clustered around the mean.
  • Grouped data and distributions (when you have many observations)
    • When data are numerous, we summarize with a distribution (a frequency table) rather than listing every value.
    • Distribution basics:
    • Class interval (or class): a range of values; e.g., 66–68, 69–71, 72–74.
    • Frequency (f): the number of observations in each class.
    • Lower limit (Li) and upper limit (Ui): the inclusive bounds of each class (for discrete data).
    • Class width: typically the size of the interval; for 66–68 this width is 3.
    • Boundary (class boundary): the value that separates two adjacent classes; e.g., between 68 and 69, the boundary is 68.5 for a continuous interpretation.
    • Example of boundaries (illustrative): if one class ends at 68 and the next starts at 69, the boundary is 68.5; similarly, boundaries might be 59.5 and 74.5 in other class layouts to enclose data points.
    • What is lost with grouped data: you only know that there are certain numbers within a class, not the exact values of individuals within that class. To know exact values, you would need the raw data.
  • Practical implications and real-world relevance
    • Different data summaries are used depending on data size and purpose (raw data vs grouped distribution).
    • Coefficient of variation helps compare variability across products or groups with different price points, scales, or units.
    • Visual aids (e.g., graphs) are valuable: a data set with large variation will look more spread out than one tightly clustered around the mean; a picture can convey dispersion quickly.
    • When communicating statistics to consumers or decision-makers, explain both the mean and the dispersion (e.g., using CV) to avoid misinterpretation of “average” reliability or duration.
  • Note on the transcript content
    • The lecture emphasizes that computing by hand can be slow and introduces shortcut formulas to speed up calculation.
    • The transcript also discusses grouped data, class intervals, and the concept of class boundaries, highlighting how grouping affects precision and the kind of information retained.
    • The discussion includes practical examples (light bulbs, prices, and consumer interpretation) to motivate why variation matters in real-world decisions.
    • Some numerical examples in the transcript appear inconsistent or mis-stated; the key concepts and formulas above remain the standard, well-accepted approach in statistics.