HHE 450: Chapter 12 & 13

What are the (2) Types of Epidemiological Error Sources?

1) Random

2) Systematic

What are the (3) Types of Systematic Error?

1) Selection Bias

2) Information Bias

3) Confounding Variables

What are the (6) Types of Confounding Variable Methods?

1) Randomization

2) Restriction

3) Matching

4) Standardization

5) Stratification

6) Multivariate Analysis

What is Systematic Error? What is it Caused by?

A persistent error that cannot be attributed to chance; leading to incorrect estimation of association.

This type of error is caused by the investigator or study participants within design and/or conduct of the study.

What is Random Error?

A type of error that arises from chance leading to an incorrect estimation of association.

Chance can be defined as an uncontrollable force with no apparent cause that arises due to unforeseeable and unpredictable processes.

True or False: Random Error Occurs in All Types of Epidemiological Studies

True!

What Causes Random Error?

Measurement Error & Sampling Variability

What is Sampling Variability?

The idea that study subjects are a sample from the population of interest.

Less Sampling Variability is Desired!

What is Statistical Inference?

A process of making statements about a population based on information from a single sample.

Accuracy of these statements depend on the validity of your sample!

What is Validity?

A Measure of Overall Accuracy of a Test

Internal Validity: Is your study measuring what it is intended to measure?

External Validity: Is it generalizable to the population?

When does Random Error Occur?

This occurs when selected subjects do not represent the underlying population just by chance.

True or False: Measurement Error can be Systematic or Random

True!

True or False: Sampling Error Causes Issues with Study Generalizability?

True!

What is Precision?

A Lack of Random Error

A state of being precise or exact!

True or False: Precision ≠ Accuracy

True!

How Can Random Error be Reduced?

By Increasing Precision

Use of accurate measurement tools, repeating measurements, increasing sample size, use of valid & reliable measurement tools

Accuracy & Precision: What Does This Depict?

Precise But Not Accurate

Small Random Error & Large Systematic Error

Accuracy & Precision: What Does This Depict?

Both Accurate & Precise

Small Random Error & Systematic Error

Accuracy & Precision: What Does This Depict?

Neither Accurate Nor Precise

Large Random Error & Systematic Error

Accuracy & Precision: What Does This Depict?

Accurate But Not Precise

Large Random Error & Small Systematic Error

How is Random Error Assessed?

Probability-Based Assessments

What are the (2) Main Approaches in Evaluating Random Error?

1) Hypothesis Testing (P-Values)

2) Confidence Intervals

What Kind of Sampling Does Probability-Based Statistics Rely on in Order to Reduce Error?

Random Sampling

What is Hypothesis Testing?

A widely used process of deciding whether to reject or not reject a hypothesis.

What is the (3) Step Process of Hypothesis Testing?

1) Identify Null & Alternative Hypothesis

2) Utilize Statistical Testing to Determine Compatibility of Study Results

3) Determine to Reject or Not to Reject the Null Hypothesis (Based on P-Value)

What is a P-Value?

If the null hypothesis is stated true, the probability of observing your study result, or something more extreme, is by chance alone.

This is used to determine compatibility of study results with the null hypothesis!

True or False: P-Values Only Evaluate Random Error & Range From 0-1

True!

What Does This Mean? (P ≤ 0.05)

This indicates a low degree of compatibility between the null hypothesis of no association and the observed result.

Small P-Value & Statistically Significant & Result is Unlikely Due to Chance

What Does This Mean? (P > 0.05)

This indicates a high degree of compatibility between the null hypothesis of no association and the observed result.

Large P-Value & Not Statistically Significant & Result is Likely Due to Chance

What Are the (2) Things to Remember When Interpreting P-Values?

1) Chance Cannot be Excluded, Regardless of How Small the P-Value & Chance Cannot be Mandated, Regardless of How Large the P-Value

2) Statistical Significance ≠ Casual Association

Your P-Value is NOT Used to Interpret These (5) Things!!!!

1) NOT the Probability the Null Hypothesis is True

2) NOT Used to Prove the Null Hypothesis is True

3) NOT Used to Prove an Exposure Causes a Disease

4) NOT the Only Number you Need to Report

5) NOT Used to Imply Medical, Biological, or Public Health Significance

Which Hypothesis do Hypothesis Tests Assume is True?

The Null Hypothesis

What (2) Statistics Are Mixed Together to Create P-Value?

1) Magnitude of Association

2) Sample Size

What are the (3) Limitations of P-Values?

1) P-Values Are Confounders

2) P-Values Do Not Rule Out Bias

3) Over-Reliance on 0.05 Cut-Point Leads to Errors

True or False: P-Value is Only (1) Factor That Should be Considered When Interpreting Study Results.

True!

True or False: Random Error Can be Caused by Measurement Error & Sampling Variability.

True!

What is a Null Hypothesis (H0)?

No Association Occurs Between Exposure & Disease

What is an Alternative Hypothesis (HA)?

Association Occurs Between Exposure & Disease

What is Point Estimate?

An observed measure of association given by study data.

Exhibits variability that can be expressed mathematically.

Wider Confidence Interval = ? Random Error/Precision

Larger Amount of Random Error & Less Precise

Narrower Confidence Interval = ? Random Error/Precision

Smaller Amount of Random Error & More Precise

What is a Confidence Interval?

A range of values within which the true magnitude of association lies with a stated probability given our observed data and assuming no bias or confounding exist.

Serves as an alternative approach to the p-value for assessment of random error.

What is Formal Statistical Definition?

The idea that if you conducted a study 100 times and obtained 100 point estimates and 100 confidence intervals, in 95 of the 100 results, the true
point estimate would lie within the given interval.

Typically 95%

True or False: Most Epidemiologists Prefer Confidence Intervals Over P-Values

True!

Why Are Confidence Intervals Preferred Over P-Values?

They are not confounded & they do a better job of separating magnitude from sample size.

What is the Correlation Between Width & Precision?

The most precise association has the smallest interval width, while the least precise association has the largest interval width.

What (3) Things Must Researches Assess to Determine if a Study is Valid?

1) Bias

2) Confounding

3) Random Error

What is Effect Measure Modification?

A variable that determines the strength of the association between an exposure and a disease.

This variable does not distort the data like confounders!

The overall goal is to describe & report rather than control it!

True or False: Confounding Must be Eliminated While Effect Measure Modification Only Has to be Explained

True!

True or False: Effect Measure Modification is NOT a Type of Bias

True!

What is an Effect Modifier?

A stratification variable that is present when stratum-specific estimates are meaningfully different rom one another.