seine: Semiparametric Ecological Inference

Ecological inference (EI) is the statistical problem of learning individual-level associations from aggregate-level data. Without certain identifying assumptions and proper estimation methods, researchers can easily draw incorrect conclusions from aggregate data, finding patterns where none exist, missing important individual-level patterns, or even concluding an effect runs in the opposite direction than it actually does. This is known as the ecological fallacy.

The seine package allows researchers to perform modern ecological inference quickly, accurately, and transparently.

• Double/debiased machine learning allows for controlling for confounding covariates, which increases the plausibility of identifying assumptions. Machine learning can be used to estimate the key regression model and avoid strong parametric assumptions made by existing EI methods.
• Sensitivity analysis and benchmarking let researchers understand how violations of their assumptions will affect results.
• A tidy interface makes the package modular and easy to use, and works well with pipe-based workflows.
• Minimal dependencies and efficient estimation routines keep everything fast and lightweight.

This software is experimental

seine and the underlying statistical methods are under active development. The package is not currently recommended for non-expert use.

Installation

You can install the development version of seine with

remotes::install_github("CoryMcCartan/seine")

Example

The package contains county-level data from the 1968 U.S. presidential election for southern states, where Hubert Humphrey (Democrat) ran against Richard Nixon (Republican) and George Wallace (independent; segregationist).

In this data, one might be interested in estimating vote choice by race. To do so, seine provides both a formula interface and a tidy interface through a new ei_spec() object. Here, we’ll demonstrate the latter.

library(seine)
data(elec_1968)

spec = ei_spec(
    elec_1968, 
    predictors = vap_white:vap_other,
    outcome = pres_dem_hum:pres_ind_wal, 
    total = pres_total,
    covariates = c(state, pop_city:pop_rural, farm:educ_coll, 
                   inc_00_03k:inc_25_99k)
)

m = ei_ridge(spec)
rr = ei_riesz(spec, penalty = m$penalty)

ei_est(regr = m, riesz = rr, data = spec, conf_level = 0.95)
#> # A tibble: 9 × 6
#>   predictor outcome estimate std.error conf.low conf.high
#>   <chr>     <chr>      <dbl>     <dbl>    <dbl>     <dbl>
#> 1 white     dem_hum   0.225     0.0241   0.178     0.273 
#> 2 black     dem_hum   0.584     0.0601   0.467     0.702 
#> 3 other     dem_hum   2.92      0.744    1.47      4.38  
#> 4 white     rep_nix   0.435     0.0365   0.363     0.506 
#> 5 black     rep_nix  -0.0242    0.0367  -0.0962    0.0477
#> 6 other     rep_nix  -4.75      0.991   -6.69     -2.81  
#> 7 white     ind_wal   0.339     0.0197   0.300     0.377 
#> 8 black     ind_wal   0.437     0.0436   0.351     0.522 
#> 9 other     ind_wal   2.84      0.840    1.20      4.49

This workflow is explained in more detail on the Get Started page, along with demonstrations of data preprocessing and sensitivity analysis.

Name

In addition to being an acronym for “semiparametric ecological inference,” seine (pronounced SAYN) refers to a type of fishing net, which is used to catch fish like tuna in aggregate. A net also is visually similar to the tomography lines which are commonly used to visualize the interaction of the latent data and the accounting identity in 2×2 ecological inference.