The Deductive Approach to Causal Inference

Pearl, Judea

doi:10.1515/jci-2014-0016

Cited by 22 publications

(11 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We illustrate Hitman through an example of an inconsistent mediator ( Figure 2). This figure is reminiscent of Simpson's Paradox (Pearl 2014b). The exposure increases the outcome, and it increases the mediator.…”

Section: Examplementioning

confidence: 99%

Novel mediation analysis of human plasma proteome and metabolome reveals mediators of improved glycemia after gastric bypass surgery

Dreyfuss

Yuchi

Hui

et al. 2019

Preprint

View full text Add to dashboard Cite

Molecular mechanisms by which Roux-en-Y gastric bypass (RYGB) improves glycemic control and metabolism in type 2 diabetes (T2D) remain incompletely understood. In the SLIMM-T2D trial, participants with T2D were randomized to RYGB or nonsurgical management and their fasting plasma proteome and metabolome were analyzed for up to 3 years. To identify analytes that mediate improvement in outcomes, we developed a high-throughput mediation analysis method (Hitman), which is significantly more powerful than existing methods. Top-ranking analyte mediators of glycemia improvement were growth hormone receptor and prolylhydroxyproline, which were more significant than any clinical mediator, including BMI. Betaalanine and Histidine Metabolism (both including CNDP1) were top differentially regulated pathways, and Valine, Leucine and Isoleucine Degradation was also a top differentiallyregulated pathway and a top mediator of improvement in insulin resistance. The identified analytes may serve as novel targets for T2D therapy. More broadly, Hitman can identify analyte mediators of outcomes in randomized trials for which high-throughput data are available. Results ClinicalProteomic and metabolomic data at baseline, the 3 month time point, and 1 year were available from 35 of the 38 SLIMM-T2D participants. Metabolic characteristics of these 35 participants did not differ between groups at baseline ( Table S1). The HbA1c and BMI of those that had proteomics or metabolomics per time point also did not differ from the SLIMM-T2D participants within their group at any time point (Figure S1). After 3 years, no DWM participants achieved study-defined glycemic goals (HbA1c<6.5% and fasting plasma glucose<126 mg/dL), whereas eight RYGB participants did, and seven of these eight participants were not receiving any antidiabetes medications (Simonson et al., 2018). The number of participants taking each diabetes medication class per arm at each time point is tabulated in Table S1. ProteomeFasting proteomics were profiled at baseline, the 3 month time point, and years 1, 2 and 3 in RYGB and DWM. At baseline (pre-randomization), there were no statistically significant proteins as assessed by false discovery rate (FDR<0.15). For later, post-randomization time points, reduction in BMI was greater in RYGB, so groups were compared both with and without adjustment for each person's BMI change. Differences in the baseline-corrected proteome in RYGB vs. DWM (i.e. differences between groups in changes over time) emerged at the 3 month time point, with 14 significant proteins in the unadjusted analysis and 8 in the BMI-adjusted analysis. Proteins common to both analyses and downregulated in RYGB were: carnosine dipeptidase 1 (CNDP1, also known as Beta-Ala-His dipeptidase 1) and Fetuin-B (FETUB). Proteins in common upregulated in RYGB were: integrin-binding sialoprotein (IBSP), insulin-like growth factor binding protein (IGFBP2), endothelial cell-specific molecule 1 (ESM1), macrophage metalloelastase (MMP12), alpha-1-antichymotrypsin complex (SERPINA3), ...

show abstract

Section: Examplementioning

confidence: 99%

Novel mediation analysis of human plasma proteome and metabolome reveals mediators of improved glycemia after gastric bypass surgery

Dreyfuss

Yuchi

Hui

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Indirect effect is to measure the portion of the effect which can be explained by mediation alone, while inhibiting the capacity of Y to respond to X . 41 The total effect is equal to the summation of direct and indirect effects.…”

Section: Effect Decomposition and Estimationmentioning

confidence: 99%

A new statistical framework for genetic pleiotropic analysis of high dimensional phenotype data

Wang

Rahman

et al. 2016

BMC Genomics

View full text Add to dashboard Cite

BackgroundThe widely used genetic pleiotropic analyses of multiple phenotypes are often designed for examining the relationship between common variants and a few phenotypes. They are not suited for both high dimensional phenotypes and high dimensional genotype (next-generation sequencing) data.To overcome limitations of the traditional genetic pleiotropic analysis of multiple phenotypes, we develop sparse structural equation models (SEMs) as a general framework for a new paradigm of genetic analysis of multiple phenotypes. To incorporate both common and rare variants into the analysis, we extend the traditional multivariate SEMs to sparse functional SEMs. To deal with high dimensional phenotype and genotype data, we employ functional data analysis and the alternative direction methods of multiplier (ADMM) techniques to reduce data dimension and improve computational efficiency.ResultsUsing large scale simulations we showed that the proposed methods have higher power to detect true causal genetic pleiotropic structure than other existing methods. Simulations also demonstrate that the gene-based pleiotropic analysis has higher power than the single variant-based pleiotropic analysis. The proposed method is applied to exome sequence data from the NHLBI’s Exome Sequencing Project (ESP) with 11 phenotypes, which identifies a network with 137 genes connected to 11 phenotypes and 341 edges. Among them, 114 genes showed pleiotropic genetic effects and 45 genes were reported to be associated with phenotypes in the analysis or other cardiovascular disease (CVD) related phenotypes in the literature.ConclusionsOur proposed sparse functional SEMs can incorporate both common and rare variants into the analysis and the ADMM algorithm can efficiently solve the penalized SEMs. Using this model we can jointly infer genetic architecture and casual phenotype network structure, and decompose the genetic effect into direct, indirect and total effect. Using large scale simulations we showed that the proposed methods have higher power to detect true causal genetic pleiotropic structure than other existing methods.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-016-3169-1) contains supplementary material, which is available to authorized users.

show abstract

“…The unmeasured covariates can be broken down into those factors that are unobserved (E ti ) and those that are unobservable (ǫ ti ) at the given level of measurement. To generalize the typical mediation problem (Pearl (2014a), Pearl (2014b), we consider a class of four DGPs in which both observed and unobserved covariates might be determined by treatment. Where variables U are unmeasured variables, the four DGPs are characterized by the following structural equations:…”

Section: Data Generating Processesmentioning

confidence: 99%

“…Following Pearl (2014a), this definition is made at the population level, with individual-level effects given by the expressions under the expectation. Expectations are taken over…”

Section: Defining Causal Effects As Changes From Interventions To a Dgpmentioning

confidence: 99%

Covariates and Causal Effects: The Problem of Context

Aliprantis

2015

Working Paper (Federal Reserve Bank of Cleveland)

View full text Add to dashboard Cite

The Deductive Approach to Causal Inference

Cited by 22 publications

References 31 publications

Novel mediation analysis of human plasma proteome and metabolome reveals mediators of improved glycemia after gastric bypass surgery

Novel mediation analysis of human plasma proteome and metabolome reveals mediators of improved glycemia after gastric bypass surgery

A new statistical framework for genetic pleiotropic analysis of high dimensional phenotype data

Covariates and Causal Effects: The Problem of Context

Contact Info

Product

Resources

About