Skip to main content

Notes on Inference

CausalImpact: Building Counterfactuals with Bayesian Structural Time Series
causal-inference
quasi-experimental-design
statistics
R

Estimate causal effects with Bayesian structural time series, examine how CausalImpact constructs the counterfactual, and compare its assumptions with DiD and synthetic…

Nov 11, 2026

The Synthetic Control Method (SCM): From Choosing a Control to Estimating Weights
causal-inference
quasi-experimental-design
statistics
R

Build a synthetic control from data-driven donor weights, evaluate its pre-treatment fit, and quantify uncertainty with in-space placebo inference.

Nov 4, 2026

Estimating the Effect with Difference-in-Differences (DiD)
causal-inference
quasi-experimental-design
statistics
R

Estimate a causal effect with difference-in-differences, assess the parallel trends assumption, and see how violations bias the result.

Oct 28, 2026

How Close Can a Pre-Post Comparison Get to the Truth?
causal-inference
quasi-experimental-design
statistics
R

Design a synthetic panel with known ground truth and use it to show why a simple pre-post comparison can mistake time variation for a causal effect.

Oct 21, 2026

Integrating Ghost Ads into an Ad-Supported Video Platform
causal-inference
incrementality
advertising
system-design

A design exploration of Ghost Ads for AVOD, connecting ad selection, control-group measurement, and incremental conversions beyond attribution.

Oct 7, 2026

Estimating Cross-Platform Usage Structure from Large-Scale Marginal Distributions and a Small Panel
audience-modeling
data-fusion
statistics
R

Fuse large-scale service logs with a small single-source panel using a Gaussian copula to estimate cross-media audience overlap and complementarity.

Sep 23, 2026

Estimating unique users for genre combinations with latent class analysis
audience-modeling
mixture-models
statistics
R

Use latent class analysis to approximate viewing patterns and estimate unique audience sizes for any combination of genres.

Sep 16, 2026

Modeling Ad Frequency Distributions with Finite Mixtures of Negative Binomials
audience-modeling
reach-frequency
statistics
R

Model heterogeneous ad exposure counts with a finite mixture of negative binomial distributions and derive reach curves from the fitted parameters.

Sep 9, 2026

Differential Attrition in Brand Lift: Diagnosing Selection Bias with Lee Bounds
brand-lift
statistics
causal-inference
R

Diagnose selection bias from differential survey response in Brand Lift studies using Lee bounds, with binary-outcome examples and reusable R code.

Aug 26, 2026

From Risk Ratios to Odds Ratios: Choosing an Aggregation Scale for Brand Lift
brand-lift
statistics
meta-analysis

Compare risk ratios and odds ratios across Brand Lift studies, derive the log-odds confidence interval, and pool campaign effects by inverse variance.

Aug 12, 2026

Fieller’s Confidence Interval for Risk Ratios: Removing the Taylor Approximation
brand-lift
statistics
R

Derive Fieller’s confidence interval for Brand Lift risk ratios, compare its coverage with delta and bootstrap methods, and implement it in R and BigQuery.

Jul 29, 2026

How Reliable Is the Delta Method CI? Bootstrap Validation and Sample Size Planning
brand-lift
statistics
R

Validate delta-method confidence intervals for Brand Lift risk ratios by simulation and bootstrap, then derive sample sizes for a target power.

Jul 15, 2026

Confidence Intervals for Brand Lift: The Delta Method on Risk Ratios
brand-lift
statistics
R

Derive a confidence interval for Brand Lift risk ratios with the log-scale delta method, including worked examples and production-ready R and BigQuery SQL.

Jul 1, 2026
No matching items
R community: R-bloggers