KM

Week-8-Common-Statistical-Techniques

Week 8: Common Statistical Techniques

Objectives for Today

  • Check-in

  • Lecture

  • Break

  • Class activity

  • Reconvene to go over class activity

Introduction to Common Statistical Tests

  • Focus on describing common tests

  • Preparation for next week's topic: interpretation of these tests

Considerations for Statistical Tests

  • Number of groups being compared

  • Groups: related (paired) or independent

  • Size of sample

  • Distribution of data

Parametric vs Non-parametric Tests

  • Parametric Tests:

    • Assume data follows a specific distribution (usually normal)

    • Sensitive to underlying distribution and sample size

  • Non-parametric Tests:

    • Fewer assumptions about data distribution

    • Used for non-normally distributed data, ordinal data, or when parametric conditions aren’t met

  • Key Differences:

    • Parametric: Assumptions about distribution; sensitive to outliers

    • Non-parametric: Robust against outliers; also known as "distribution free"

When to Use Parametric Tests

  • When data meets assumptions: normality and sample size

  • Large sample size: >25

  • Continuous outcome variables

When to Use Non-parametric Tests

  • Small sample size: <25

  • Heavily skewed data

  • Categorical outcome data

Independent vs. Dependent Samples

  • Independent Samples:

    • Unrelated groups (e.g., comparing blood pressure in nurses vs doctors)

  • Dependent Samples:

    • Related groups (e.g., blood pressure before and after drug treatment in the same patients)

Common Statistical Tests

  • Chi-square

  • T-test

  • Correlation

  • Regression

Chi-square Test

  • Null Hypothesis (Ho): No relationship between categorical variables

  • Alternative Hypothesis (Ha): Relationship exists

  • Non-parametric:

    • Independent samples

    • Outcome: nominal or ordinal

  • McNemar test for paired data

Chi-square Methodology

  • Based on counts, not standard deviation

  • Compares expected vs. observed frequencies

  • Does not measure strength of association

Examples for Chi-square Use

  • HPV vaccination status and cervical cancer relationship

  • Lung cancer and smoking status correlation

Correlation Tests

  • Ho: No correlation between two variables

  • Ha: There is a correlation

  • Focus on the strength and direction of the relationship

  • Pearson's r: Measures linear relationships

  • Spearman correlation: Non-parametric equivalent

Examples for Correlation Use

  • Correlation between height and age

  • Steps per day and age at mortality

T-test (Independent)

  • Ho: No significant difference between means of two groups

  • Ha: Significant difference exists

  • Binary independent variable with continuous outcome

  • Mann-Whitney U test: Non-parametric equivalent

Examples of T-test Use

  • Comparing V02 max in intervention groups

  • Audiogram scores vs. noise exposure

ANOVA (One-way)

  • Ho: No significant difference between means of three or more groups

  • Ha: Significant difference exists

  • Categorical independent variable with continuous outcome

  • Kruskal-Wallis test: Non-parametric equivalent

Examples for ANOVA Use

  • Fall risks among different age groups

  • Patient readmission rates between hospitals

Paired T-test

  • Ho: No difference between means of paired data

  • Ha: Significant difference exists

  • Compares data from same subjects under different conditions

  • Wilcoxon signed-rank test: Non-parametric equivalent

Examples for Paired T-test Use

  • Knowledge scores before and after an education intervention

  • Mean A1c levels before and after diabetes intervention

Importance of Understanding Statistical Tests

  • Helps interpret results and understand research methodologies

  • Basic knowledge is crucial for analyzing studies

Reminders for Next Week

  • Questions or concerns?

  • Assignment 1 due November 8, 5 PM: Group work required; one submission needed per group, all submit quiz portion.