Time Series Designs in Research Methods

Time Series Designs

• A research method using repeated measurements over time to observe changes due to an intervention or event.
• Pros: Tracks changes over time, useful for natural experiments/policy changes, reduces threats to internal validity (when well-designed).
• Cons: Sensitive to external confounding events, requires consistent reliable data collection, may struggle with causal inference without proper controls.

Comparison Groups

• Adding a comparison group improves causal inference by highlighting changes unrelated to the intervention.
• A comparison group always helps to increase internal validity of the design.
• Propensity score matching: A statistical technique to match individuals in the treatment and comparison groups.

Selection Bias

• Occurs when groups differ systematically, causing outcomes to be attributed to pre-existing differences instead of the intervention.

Simple Interrupted Time Series Design

• Measures outcomes before and after an intervention/event.
• Pros: Simple, intuitive, shows clear trends.
• Cons: No control for external factors, changes might not be due to intervention.

Interrupted Time Series Design with Reversal

• Measures outcomes before, during, and after an intervention reversal (reintroducing the policy).
• Pros: Demonstrates causality by observing if reversal undoes effects.
• Cons: Reversals may not always occur or be feasible.

Interrupted Time Series Design with Multiple Replications

• Repeatedly introduces and removes an intervention to observe consistent effects.
• Pros: Strong causal inference.
• Cons: Requires careful timing and replication.

Comparison Group Interrupted Time Series Design

• Adds a non-treated comparison group to observe natural changes versus intervention-induced changes.
• Pros: Controls for external factors affecting both groups.
• Cons: Requires well-matched comparison groups.

Regression Discontinuity Design

• Assigns treatment based on a cutoff score, examining outcomes above and below the cutoff.
• Pros: High internal validity for causal inference, doesn't require random assignment
• Cons: Only works near the cutoff (limited generalizability), requires a sharp cutoff and reliable assignment mechanism.

Correlational Designs

• Examines the relationship between two or more variables without manipulation.
• Correlation does not equal causation.

Correlation Coefficients

• Pearson's correlation coefficient (r) measures the strength and direction of a linear relationship between two variables.
• Magnitude: Values range from -1 to 1, indicating strength of association
• Direction: Positive (r > 0) or negative (r < 0)
• Significance: P-value less than 0.05 indicates statistical significance.
• r^2 (Coefficient of Determination): Proportion of variance in one variable explained by the other.

Simple Linear Regression

• Summarizes how predictor variables are related to the outcome variable.
• Allows prediction of the outcome based on predictor variables.
• Assumes a linear relationship between variables.

Multiple Regression

• Predicts a dependent variable using multiple independent variables.
• This is useful for including multiple predictors to better predict an outcome
• Includes aspects like: Standard (simultaneous), stepwise, hierarchical.

Hierarchical Regression

• Enters predictors into blocks based on theoretical or research priorities
• Aims to test if new predictors add unique variance beyond covariates or earlier blocks.
• Key statistics for interpretation include R^2 change, Regression F-test, unstandardized/standardized coefficients (b/Beta), and significance.

Moderation

• Occurs when the relationship between a predictor (X) and an outcome (Y) changes depending on the level of a third variable (moderator, M).
• Implies that the relationship between two variables is different at different levels of a third variable
• Indicates whether the effect of one variable on another depends on the level of a third variable
• Used to model interaction effects.

Mediation

• Occurs when the relationship between a predictor (X) and an outcome (Y) is explained (fully or partially) by an intermediary variable (mediator, M).
• Suggests that the effect of a predictor is exerted through a mediator
• Used to model the process by which a predictor influences an outcome.

Setting Level Research

• Examines the influence of environments/contexts on outcomes.
• Focuses on how characteristics of a setting shape individual or group outcomes.
• Types of settings include proximal factors (influences close to individuals), distal factors (broader influences).
• Tools used in setting-level research include cluster-randomized control trials & correlational research.
• Multilevel modeling is used to analyze data involving nested structures (e.g., students within schools).

Community-Engaged Research

• Philosophy that is rooted in community expertise and collaborative practices
• Types of approaches include community-based participatory research (CBPR) & participatory action research (PAR), which emphasises co-creation of knowledge, and mutual benefit.
• Approaches may include community involvement in data collection, analysis, and recommendations, ensuring the project is culturally relevant and sustainable while respecting the communities ethical practices and principles.