Get started with MetaHunt

What this is

MetaHunt is a privacy-preserving meta-analysis tool. You start with per-study estimated functions and a row of metadata for each study, and end with a predicted function (or scalar summary) for a new study, together with a conformal prediction interval. This vignette shows the shortest possible end-to-end run; for the assumptions, the three-step pipeline, and tuning, see vignette("metahunt-intro", package = "MetaHunt").

Inputs in 30 seconds

Two ingredients drive every function in the package.

Starting from fitted models? If your inputs are one fitted model per study (e.g. ranger, grf::causal_forest, lm), turn them into F_hat first with f_hat_from_models() after building a shared evaluation grid via build_grid(). See vignette("data-prep", package = "MetaHunt") for a worked example.

A minimal end-to-end run

Simulate a small problem so the vignette has no external data dependency:

m <- 30; G <- 20; K_true <- 2
x <- seq(0, 1, length.out = G)
basis <- rbind(sin(pi * x), x)                        # 2 true bases on the grid

W <- data.frame(w1 = rnorm(m), w2 = rnorm(m))
beta <- cbind(c(1, -0.5), c(-0.5, 1))
pi_true <- exp(as.matrix(W) %*% beta)
pi_true <- pi_true / rowSums(pi_true)

F_hat <- pi_true %*% basis + matrix(rnorm(m * G, sd = 0.05), m, G)
W_new <- data.frame(w1 = c(0, 1, -1), w2 = c(0, -0.5, 1))  # three new metadata profiles

Fit and get a conformal interval for the scalar wrapper mean (the ATE under uniform grid weights):

res <- split_conformal(F_hat, W, W_new, K = K_true,
                       wrapper = mean, alpha = 0.1, seed = 1,
                       dfspa_args = list(denoise = FALSE))
data.frame(prediction = res$prediction,
           lower      = res$lower,
           upper      = res$upper)
#>   prediction     lower     upper
#> 1  0.5510758 0.5239399 0.5782117
#> 2  0.6033022 0.5761663 0.6304381
#> 3  0.5003344 0.4731985 0.5274703

Visualising the result

plot() on the conformal object renders the prediction with its band:

plot(res)

Where to next

• vignette("metahunt-intro", package = "MetaHunt") — the four assumptions, the three-step pipeline, choosing K, denoising, conformal flavours, and minmax_regret().
• vignette("data-prep", package = "MetaHunt") — turning per-centre fitted models into (F_hat, W).
• ?metahunt — the full pipeline reference.