Icon

Advanced A B Testing for Panel Data Experimentation

<p>This workflow supports advanced A B testing using panel data. It enables causal effect estimation with treatment indicators, multiple control variables, and unit and time fixed effects. The workflow is designed for experimentation scenarios such as staggered treatment adoption, difference in differences analyses.</p>

This dataset contains synthetic panel data for Difference in Differences experimentation.

The outcome variable is generated using the following model:

Y_it = alpha_i + lambda_t + beta D_it + gamma X_it + epsilon_it

where

  • i indexes the unit

  • t indexes time

  • alpha_i represents unit fixed effects

  • lambda_t represents time fixed effects

  • X_it is a set of control variables

  • D_it is the treatment indicator

  • beta is the causal treatment effect under the parallel trends assumption

The treatment indicator is defined as:

D_it = 1 if unit i is treated and time t is greater than or equal to the adoption time T_i
D_it = 0 otherwise

Under the Difference in Differences framework, the treatment effect beta is identified by comparing outcome changes over time between treated and untreated units, while controlling for unit specific heterogeneity and common time shocks.

Synthetic generic panel dataDiff in Diff RegressionYit​=αi​+λt​+βDit​+γ⊤Xit​+εit​
CSV Reader
Plot the estimated coefficients with Confidence Intervals
Python View
Each row is a unit–time observation. Columns are: unit: panel unit identifier (index 1)time: time period (index 2)Y: outcome variableD: treatment dummy (post adoption for treated units)X1, X2, X3: control variables
Linear Model Learner

Nodes

Extensions

Links