Repository with codes for the paper "Data integration of non-probability and probability samples with predictive mean matching"

Basic info

Repository for Chlebicki P., Chrostowski, Ł. & Beręsewicz, M., "Data integration of non-probability and probability samples with predictive mean matching".

• preprint @ arXiv
• How to cite: Chlebicki, P., Chrostowski, Ł., & Beręsewicz, M. (2024). Data integration of non-probability and probability samples with predictive mean matching. arXiv preprint arXiv:2403.13750.

Setup

Install the development version of the nonprobsvy package from GitHub:

install.packages("remotes")
remotes::install_github("ncn-foreigners/nonprobsvy@dev") ## 

Structure of the repo

• codes/
  • sim-run-all.R - the code to run all simulations (it takes about 8 hours with 8 cores)
  • sim-main-paper.R - main simulation study (section 4)
  • sim-appen-1-choose-k.R -- additional simulation (section C.1)
  • sim-appen-2-varsel.R -- additional simulation (section C.2)
  • sim-appen-3-nonparam.R -- additional simulation (section C.3)
  • sim-appen-4-positivity.R -- additional simulation (section C.4)
  • sim-appen-5-multirobust.R -- additional simulation (section C.5)
  • reporting.R -- code to generate tables and figures from the paper
• results/
  • sim1-paper-results.RDS -- main simulation study results (table 4.1 and 4.1)
  • sim1-paper-results-no-v2.RDS -- main simulation study results (table 4.2)
  • sim-appen1-choose-k-results.RDS -- additional simulation (section C.1)
  • sim-appen2-varsel-results.RDS -- additional simulation (section C.2)
  • sim-appen3-nonparam.RDS -- additional simulation (section C.3)
  • sim-appen4-positivity.RDS -- additional simulation (section C.4)
  • sim-appen5-multirobust.RDS -- additional simulation (section C.1)

Basic codes

For this paper we developed two predictive mean matching (PMM) estimators and implemented them into the nonprobsvy package. Below you can find functions to run the

• PMM estimator $\hat{y}-\hat{y}$ matching

PMM_A <- nonprob(outcome = y1 ~ x1 + x2,
                 data = sample_nonprob,
                 svydesign = sample_prob,
                 method_outcome = "pmm",
                 family_outcome = "gaussian",
                 control_outcome = controlOut(k = 5, predictive_match = 1),
                 control_inference = controlInf(pmm_exact_se = TRUE))

• PMM estimator $\hat{y}-y$ matching

PMM_A <- nonprob(outcome = y1 ~ x1 + x2,
                 data = sample_nonprob,
                 svydesign = sample_prob,
                 method_outcome = "pmm",
                 family_outcome = "gaussian",
                 control_outcome = controlOut(k = 5, predictive_match = 2),
                 control_inference = controlInf(pmm_exact_se = TRUE))

Funding

Work on this paper was supported by the National Science Centre, OPUS 20 grant no. 2020/39/B/HS4/00941.