module 3 stat section B 1
Module 3 Overview
Welcome to Module 3, focusing on statistical concepts from Module 2.
Module Structure
Section A: Analysis of categorical data
Part 1: Univariate analysis of categorical data
Part 2: Analysis of relationships between categorical variables
Part 3: Multivariate analysis of categorical data
Section B: Analysis of continuous data
Objectives of the Module
Determine the most appropriate statistical test for different data types and comparisons.
Understand the rationale and assumptions behind each statistical test.
Interpret and report the results of the statistical analyses.
Categorical Data Analysis
Summarizing Categorical Data:
Use frequencies, proportions, or percentages for single variables.
Employ a two-way contingency table (cross tabulation) to summarize relationships between two categorical variables.
Key Questions in Categorical Data Analysis
Is the observed proportion the same as a hypothesized proportion?
Does our sample proportion align with another study's population proportion?
Is there a difference in proportions between two or more groups?
Does an intervention change the outcome proportions?
Univariate Analysis of Categorical Data
Example Question:
"Is the proportion of first-year QUT students who are female the same as in the Brisbane population aged 18-21?"
Analysis Type: Compare sample proportion to an external population.
Another Example Question:
"Is the distribution of individuals presenting to Royal Brisbane Hospital's emergency ward evenly spread across age groups?"
Analysis Type: Compare sample results to a hypothesized distribution.
Statistical Tests for Categorical Data
Chi-Square Goodness of Fit Test
Purpose: Used to compare observed and expected frequencies for a single categorical variable.
Hypotheses:
Null Hypothesis (H0): Observed frequencies are evenly distributed across categories.
Alternative Hypothesis (H1): Observed frequencies are not evenly distributed.
Data Example: Accident Emergency Ward
Observed Frequency Data for Age Groups:
0-15: 40
16-30: 62
31-60: 50
60+: 42
Total Sample Size: 200
Expected Frequencies (Assuming H0 is true)
Each age group should have 50 individuals if evenly distributed.
Calculation of Chi-Square Statistic (χ²)
Formula: χ² = Σ[(Observed - Expected)² / Expected] for each group.
Example Calculation:
For age group 0-15:
Observed: 40, Expected: 50
(40 - 50)² / 50 = 2
Continue for all groups and sum results for total chi-square value.
Test Statistic and P-Value Calculation
A test statistic is computed to determine whether observed values support the null hypothesis.
Finding Significance: If the calculated p-value is greater than 0.05, we do not reject the null hypothesis.
In this example:
χ² value = 6.88, P-value = 0.076
Conclusion: No evidence to suggest that the age distribution is uneven; frequencies are in line with H0.
Chi-Square Distribution
Non-symmetric, shape depends on degrees of freedom.
Degrees of Freedom Formula: df = number of groups - 1.
For our example, df = 4 - 1 = 3.
Chi-square values greater than a certain threshold suggest significant results.
Why Use Software?
Calculating χ² values traditionally requires referencing tables, which can be tedious.
Using statistical software like R Commander automates calculations and improves accuracy.
Recap
Chi-square goodness of fit is used to test proportions across categories.
Hypothesized distribution example: age groups presenting evenly in emergency studies.
Next Steps
Future topics will cover bivariate categorical data analysis.