Overview | Which inference method should I use? | 1. Build the synthetic control | 2. Prediction intervals: synth_inference() | Split-conformal (default) | Parametric (Gaussian residuals) | When to use which | 3. Placebo inference | Generate the placebos | Test | Plot the placebo distribution | Plot the placebo gaps | 4. Combining prediction intervals and placebos | Alternative QP backends | See also | References

1. Build the inputs | 2. Fit the synthetic control | 3. Inspect | 4. What next?

Motivation | A scaling comparison | Scaling beyond what exact KRLS can do | Reusing landmarks across fits | Choosing the lambda criterion | What carries over from the exact path | Limitations

A toy panel | Three fits | Where the weight goes | Diagnostics | Plots | Choosing an estimand

A simulated example | Fit | Weighted lm() with robust SEs | Adding regression adjustment (doubly-robust) | Survey-style inference | Trimming if weights blow up | Choice of estimand for inference

What is entropy balancing? | A minimal example | Fit | Tidy output | Plotting | Using the weights downstream | Two solver methods | See also | References

Why KRLS? | A simulated example | Fit | Tidy output | Visualize | Augment for downstream analysis | When KRLS is the right tool | See also | References