Identification and Estimation of the Bi-Directional MR with Some Invalid Instruments
Authors: Feng Xie, Zhen Yao, Lin Xie, Yan Zeng, Zhi Geng
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we conduct various simulation studies to evaluate the performance of the proposed PRe Bi M method. |
| Researcher Affiliation | Academia | Feng Xie1, Zhen Yao1, Lin Xie1, Yan Zeng1, , Zhi Geng1 1Beijing Technology and Business University |
| Pseudocode | Yes | Algorithm 1 PRe Bi M |
| Open Source Code | Yes | Our source code can be found in the Supplementary Materials. |
| Open Datasets | Yes | We first apply our method to analyze the bi-directional causal relationships between obesity and vitamin D status using the GWAS data from Vimaleswaran et al. [2013]. Next, we apply our method to analyze the causal effect of institutions on economic development using the Colonial Origins dataset from [Acemoglu et al., 2001]. |
| Dataset Splits | No | T1: Sensitivity to Sample Size. We evaluated the impact of different sample sizes: n = 2k, 5k, and 10k, where k equals 1, 000. |
| Hardware Specification | Yes | All experiments were conducted using AMD Ryzen 7 7735H with Radeon Graphics processors, operating at a base speed of 3.20 GHz, and equipped with 16.0 GB (15.2 GB available) of RAM. |
| Software Dependencies | No | For sis VIVE algorithm, we used the implementations in the R sis VIVE package, which can be downloaded at https://cran.r-project.org/web/packages/sis VIVE/. |
| Experiment Setup | Yes | To compare the performance of these methods in a realistic setting, analogous to Slob and Burgess [2020], the genetic variants are modeled as Single Nucleotide Polymorphisms (SNPs), with a varying minor allele frequency maf j, and take values 0, 1, or 2. The minor allele frequencies are drawn from a uniform distribution. Specifically, the data generation process for the bi-directional model is as follows: U = G γU + ε1, (11) X = Y βY X + G γX + UγX,U + ε2, (12) Y = XβX Y + G γY + UγY,U + ε3, (13) Gij Binomial(2, maf j), maf j U(0.1, 0.5), (14) where the error terms ε1, ε2, ε3 each follow an independent normal distribution with mean 0 and unit variance. The causal effects βY X and βX Y are generated from a uniform distribution between [ 1, 0.5] [0.5, 1]. |