Descriptive and Bivariate Statistics

Descriptive Statistics

• Standard Deviation (S.D. or o)
  • Average deviation of each observation from precise centre of all weighted observations, but sensitive to relative weights of observations.
    • Indicates how dispersed points are from the mean
    • More precise than mean deviation; preferred.
• Example:
  • David Demers….
    • N = 198 newspapers
    • M = 2.64 editorials
    • o = +/- 1.84
    • Range = 0 – 8
    • Standard deviation tells us that the amount of editorials/paper is not around 2.64, but between the dispersion of 0.8 and 4.48 (2.64 +/- 1.84)

Bivariate Statistics

• Analysis of two sets of data
• Coefficient of Correlation
  • Indicates extent to which ordered pairs of observations share variance
  • It gauges if the behaviours of two phenomena are related.
    • Example
    • If x happens, does y happen?
    • If x happens, how much does y happen?
• For correlation and scatter points:
  • Ordered pair = two related numbers (x, y)
  • Examples:
    • (x,y) [scatter plot of x and y analysis, use excel]
      • Hours of television, hours of radio
      • Age, hours of television
      • Education (coded?), hours of radio
      • Age, education.
• Coefficient of correlation or Pearsons r
  • “Coefficient” = measure of strength of correlation expressed as r = + z or – z
    • 0-10 (no correlation)
    • 10-30 (slight degree of relationship)
    • 30-50 (fair degree of relationship)
    • Common knowledge…
• Descriptive Statistics àCorrelation of Coefficient
  • Can do
    • Flag significant data (e.g. highs, lows, averages)
    • Reveals patterns
    • Suggests relationships between values
    • Indicate strength relationship between two values
  • Can’t do
    • Provide explanations
    • You must provide the explanation

Cool Related Sites:

On SEOMOZ “Knowledge” about Statistics « IR Thoughts
Statistics for Business and Economics « Khmer Campus – for unde..
Language Log » Physiological politics
Data Analysis Part 1: Database Management, Distributions, and Bivariates &l..
A Little Primer on Multicollinearity « DemocraticPeace Blog
Good Government vs. Less Government « The Baseline Scenario

Category: Research Methods