Standard Deviation vs. Standard Error n = sample size Sigma (σ) = standard deviation SEM = standard error of the mean SEM = σ/√n SEM < σ SEM decreases as n increases z-scores 1 = +/- 1 σ around mean 2 = +/- 2 σ around mean 3 = +/- 3 σ around mean Confidence Interval (CI) Describes the range in which the mean would be expected to fall if the study were performed again and again = range from [mean - Z(SEM)] to [mean + Z(SEM)] 95% CI (alpha = 0.05) is standard for 95% CI, Z = 1.96 Outcomes if 0 falls within the CI when calculating the difference between 2 variables, H0 is not rejected and the result is not significant if 1 falls within the CI when calculating OR or RR, H0 is not rejected and the result is not significant Significance statistical significance refers to whether p < 0.05 clinical significance requires that a statistically significant result be also clinically meaningful T-test vs. ANOVA vs. χ2 T-test compares the means of 2 groups on a continuous variable ANOVA (analysis of variance) compares the means of 3 or more groups on a continuous variable χ2 ("chi-squared") tests whether 2 nominal variables are associated used with 2x2 tables e.g., effect of treatment on disease Correlation Coefficient (r) Pearson coefficient, r, is always between -1 and +1 Absolute value indicates strength of correlation between 2 variables Coefficient of determination = r2 Attributable Risk (AR) AR is incidence in the exposed (Ie) - incidence in the unexposed (Iu) = Ie - Iu Ie = a/(a+b) Iu = c/(c+d) AR = a/(a+b) - c/(c+d) The AR percent (ARP) is the attributable risk divided by incidence in the exposed (Ie) ARP = 100* (Ie-Iu)/Ie = 100*[a/(a+b) - c/(c+d)]/[a/(a+b)] note that relative risk (RR) = Ie/Iu = a/(a+b) DIVIDED BY c/(c+d) using math tricks ARP = (RR-1)/RR Regression Analysis Regression analysis allows for the an outcome variable to be estimated from various predictor variables while adjusting for covariates Linear regression Outcome variable is continuous Predictor variables can be continuous or categorical Logistic regression Outcome variable is binary Predictor variables can be continuous or categorical Exponentiation of the coefficients yields the odds ratio for that predictor