STAT 018: Basic Study Design

• STAT 018: Statistics Applied in Communication is offered by the Department of Mathematics and Statistics, College of Science.

Basic Study Design

• Encompasses observational studies and experimental designs.

Observational Studies

• Researchers observe subjects in their natural setting without intervening to influence outcomes.
• Investigators identify subjects as they naturally occur and record responses of interest.
• Prospective Studies: Subject's explanatory level is identified first, then the outcome is observed.
• Retrospective Studies: Subject's outcomes are observed first, then information on explanatory variables is obtained.

Types of Observational Studies

• Cross-sectional studies gather data at a specific moment or within a short period
  • They randomly select subjects to assess explanatory and response variables.
  • These studies typically use large medical databases retrospectively.
  • They are generally quick and cost-effective, but may not produce highly conclusive results.
  • Data collected over a short period may limit the study’s ability to give a complete picture as long-term effects may be missed.
• Case-control studies select subjects based on their response variable and measure their explanatory variable.
  • Individuals with a specific characteristic are compared to those without, such as patients diagnosed with a disease versus a control group.
  • Studies are typically retrospective.
• Cohort studies select subjects based on their explanatory variable and observe their response outcome over time
  • Participants are tracked over an extended period to monitor outcomes.
  • Large sample sizes are required when the outcome of interest is rare.
  • Participant dropout is a key challenge.
  • These studies are typically prospective.
  • One use is when assigning conditions would be unethical.
• Cohort studies can establish temporal relationships between variables.

Summary of Observational Studies

• Case-control studies select participants based on their outcome status.
• Cohort studies select participants based on their exposure status.
• Cross-sectional studies select participants based on predefined inclusion and exclusion criteria.

Experimental Design

• Researcher makes an intervention and the subjects are followed and measured.

Key Terms

• Experimental Unit/Subject: A person, object, or entity on which a treatment is applied.
• Factor: A controlled independent variable manipulated by the experimenter.
• Response Variable: The primary outcome or measure of interest.
• Control Group/Non-Exposed Group: Serves as a baseline treatment for comparison.
• Case/Exposed Group: A group of experimental units where the factor is applied.
• Explanatory variable: variable that we can manipulate or observe that causes changes.
• Response variable: variable that changes as a result

Blinding

• A technique used to prevent bias by ensuring subjects and/or researchers are unaware of treatment details.
• Single-Blind Study: Participants are unaware of their assigned treatment, but the researchers do know.
• Double-Blind Study: Neither participants nor researchers know which group received which treatment; a third party typically holds this information.

Steps in Designing an Experiment

• Step 1: Identify the Problem to be Solved
  • Clearly define the research question.
  • Specify the response variable and the population being studied.
• Step 2: Determine the Factors Affecting the Response Variable
  • Identify variables that may influence the response.
  • Decide which factors to control, manipulate, or leave uncontrolled.
• Step 3: Determine the Number of Experimental Units
  • Establish the sample size required, considering resource limitations.
• Step 4: Determine the Levels of Each Factor
  • The level of control is keep a factor constant if its effect is not being studied, or vary its levels if analyzing its influence on the response variable
  • Randomly assign experimental units to treatments to minimize bias.
• Step 5: Conduct the Experiment
  • Implement the study according to the planned design.
• Step 6: Test the Claim
  • Analyze the data to determine if the results support or refute the initial hypothesis.

Types of Experimental Designs

• Completely Randomized Design: Each experimental unit is randomly assigned to a treatment group.
• Matched-Pairs Design: Experimental units are paired based on similar characteristics and receive different treatments.
• Randomized Block Design: Experimental units are grouped into homogeneous blocks before random assignment of treatments.

Observational Studies vs. Designed Experiments

• Observational studies are more cost-effective and applicable to a broader range of participants.
• Designed experiments allow to draw conclusions about causation, whilst observational studies do not control variables, meaning their results can only establish associations.
• Observational studies observe associations, designed experiments can show causation.

Superior choice

• Designed experiments provide stronger causal evidence, but, conducting an experiment may be unethical or impractical, making observational studies the only viable approach.

Potential Problems

• Confounding: Occurs when the effects of two or more explanatory variables are not separated, leading to uncertainty about which variable is responsible for any observed effect.
• Lurking variable is an explanatory variable that was not considered in a study, but that affects the value of the response variable in the study.
• Confounding variable A confounding variable is an explanatory variable that was considered in a study whose effect cannot be distinguished from a second explanatory variable in the study.
• Lurking variables create the apparent association, while confounding variables do not necessarily have any association with the other explanatory variable, but does have an effect on the response variable.
• Lurking variables are not considered in the study while confounding variables are measured in the study.