Inferential Statistics

• Inferential test determines significant correlation between variables, showing impact of IV or chance, accepting alternate hypothesis if significant, rejecting null hypothesis if not
• Significance level refers to result probability due to chance factors; p < 0.05 (most): less than 5% probability results due to chance (95% confidence), p < 0.1: less than 10% (90% confidence), and p < 0.01: less than 1% (99% confidence)
• • Type I error: Alternate hypothesis supported and null rejected, but there was not a real significant correlation (level of significance too lenient - p < 0.1), false Positive
  • Type II error: Alternate hypothesis is rejected and null supported, but there was a real significant correlation (level of significance too cautious - p < 0.01), false Negative
• One-tailed test use directional hypothesis, predicting difference and direction, but two-tailed use non-directional, predicting difference but not direction (they have different critical values)
• When you do inferential test result is observed value; to test significance, it is compared to critical value for used statistical test
• Mann-Whitney U or Wilcoxon: Observed value must be under/equal to critical value, Spearman’s Rho or Chi-Squared: Observed value must be over/equal to critical value
• Inferential Testing

  • Determines significant correlation between variables
  • Show IV impact or chance
  • Accepting alternate hypothesis if significant, null if not
  • Significance level is result probability due to chance factors
  • p < 0.05 (most): > 5% probability results from chance (95% confidence)
  • p < 0.1: > 10% (90% confidence)
  • p < 0.01: > 1% (99% confidence)
  • One-tailed - directional, predict difference + direction
  • Two-tailed - non-directional, predict only difference
  • Result is observed value; compared to critical value
• Errors

  • Type I error:
  • Alternate hypothesis supported + null rejected
  • But there was no real significant correlation
  • Significance level too lenient - p < 0.1
  • False Positive
  • Type II error:
  • Alternate hypothesis rejected + null supported
  • But there was a real significant correlation
  • Significance level too cautious - p < 0.01
  • False Negative