Stats - Variation

Measures of Variation

Overview of Measures of Variation

• Measures of variation assess how spread out a dataset is.
  • Greater spread results in larger values for range, variance, and standard deviation.
  • Increased concentration of data yields smaller values for range, variance, and standard deviation.
• A dataset with no variability produces all measures equal to zero.
  • Example: Dataset = [5, 5, 5, 5, 5, 5, 5]
  • Range: Maximum - Minimum = 5 - 5 = 0
  • Variance and Standard Deviation = 0, aligning with no variation concept.

Characteristics of Variation Measures

• None of the measures can be negative; they are always zero or positive.
  • This is due to them representing an absolute measure of distance over which data points are spread.

Coefficient of Variation (CV)

• Abbreviated as CV, it assesses relative variation.
  • Relative variation refers to the measurement based on a comparative basis, often as a percentage.
• Formula for Coefficient of Variation:
  $CV = \left( \frac{\text{Standard Deviation}}{\text{Mean}} \right) \times 100$
• Importance: Enables comparison of datasets measured in different units.
  • For instance, comparing temperature data in Fahrenheit vs. Celsius.
• Example:
  • Stock A: Mean = 50, Standard Deviation = 5
  • Stock B: Mean = 100, Standard Deviation = 5
  • If only standard deviation is considered, they appear the same.
  • Relatively, stock A has a higher CV (10%) compared to stock B (5%), indicating stock B has lower variability relative to its mean.

Calculating Coefficient of Variation

• For Stock A:
  • $CV_A = \left( \frac{5}{50} \right) \times 100 = 10\%$
• For Stock B:
  • $CV_B = \left( \frac{5}{100} \right) \times 100 = 5\%$
  • This establishes that stock A is more variable relative to its mean than stock B.

Outliers

• Defined as extreme values that can significantly affect statistical calculations (mean, range, variance).
• Method to identify outliers: Z-score method.
• Formula for Z-score: $Z = \frac{X - \mu}{\sigma}$ where:
  • X = data point
  • \mu = mean
  • \sigma = standard deviation
• Steps to identify outliers:
  1. Centering: Adjust dataset to have a mean of zero.
  2. Scaling: Normalize standard deviations across the dataset.

Example of Z-score

• Given data with a mean of 150 and a standard deviation of 50, to find the Z-score for a specific data point
  • Centering:
  • Minimum = 0, Maximum = 300
  • Shift mean by subtracting 150 from each point, resulting in:
    • For data point 620:
      $Z = \frac{620 - 490}{100} = 1.3$
  • Not considered an outlier since Z-score lies between -3 and 3.

Skewness and Kurtosis

• Skewness measures asymmetry of the distribution.
  • Left skewed (negatively skewed): Mean < Median.
  • Right skewed (positively skewed): Mean > Median.
• Kurtosis measures the sharpness of the distribution peak.
  • Leptokurtic: Higher, sharper peak (positive kurtosis).
  • Platykurtic: Flatter peak (negative kurtosis).
• Interpretation of skewness and kurtosis is important; not expected to compute, but to understand implications.

Quartiles and Five-Number Summary

• Quartiles split the dataset into four segments, each containing 25% of the data.
  • Q1: First quartile, separates lower 25% from upper 75%.
  • Q2: Second quartile (median), separates lower 50% from upper 50%.
  • Q3: Third quartile, separates lower 75% from upper 25%.
• Continuous measure of position in an ordered dataset:
  • Q1's position: $\frac{n + 1}{4}$
  • Q2's position: $\frac{n + 1}{2}$
  • Q3's position: $\frac{3(n + 1)}{4}$
• Interquartile Range (IQR):
  • Measures the middle 50% of the data:
  • $IQR = Q3 - Q1$
• Five-Number Summary comprises: Minimum, Q1, Median, Q3, Maximum, providing a comprehensive view of data spread.

Relationships in Five-Number Summary

• Left skewed data: Difference between median and minimum > median and maximum.
• Symmetrical data: Differences are approximately equal.
• Right skewed data: Difference between median and minimum < median and maximum.