Post-Doctoral Position

Universities and Institutes of France

France

October 20, 2022

Description

  • Organisation/Company: INSERM - DR Occitanie Pyrénées
  • Research Field: Biological sciences › Biology Computer science › Database management
  • Researcher Profile: First Stage Researcher (R1)
  • Application Deadline: 20/10/2022 21:00 - Europe/Brussels
  • Location: France › Toulouse
  • Type Of Contract: Temporary
  • Job Status: Full-time
  • Hours Per Week: 38
  • Offer Starting Date: 02/01/2023
  • Is The Job Related To Staff Position Within A Research Infrastructure?: Yes
  • Main task

    For more than 20 years, we have seen a strong growth of the causal inference approaches in Epidemiology. These methods are usually applied for confirmatory analyses, following several steps, including: the specification of a hypothetical causal model based on our knowledge and the literature, using Directed Acyclic Graphs (DAGs); the specification of a causal quantity of interest, expressed in counterfactual form, corresponding to the scientific question; the assessment of the identifiability of the target causal quantity; the specification of a statistical model, choice of an estimator and its application using statistical software; and finally the interpretation of the results (Petersen ML et al. Epidemiology 2014;25:418-26).

    Within the “Expanse” European project, which is focused on exploring the influence of the urban exposome on health, we are studying the effect of complex and multidimensional exposures (in particular related to the social environment) on health. In order to better understand the mechanisms explaining the links between early exposures and later health status, mediation analyses can be applied to characterize the role of intermediate variables between the exposure and the health outcome. The main methodological advances in mediation analyses have been developed in the context of causal inference analyses: the decomposition of total effects into two, three, or four components, as well as a better handling of confounding, censoring or missing data. Different estimators have also been developed (g-computation, Inverse Probabiliy of Treatment Weighting, double robust estimators, etc).

    The researcher who will be hired will have the mission to work on mediation analyses that can involve a large number of mediators and time-varying covariates. Two main topics can be studied: 1) The initial exploratory approach of a mediation analysis, in the presence of many possible mediators; 2) In a more confirmatory framework, the optimization of methods to estimate direct and indirect effects using high dimensional data.

    Main activities

  • Regarding exploratory analyses, the aim is to optimize the mediation analysis approach in the presence of many mediators or in case of a limited theoretical knowledge about the causal structure of a Directed Acyclic Graph. The use of “causal discovery” methods (Glymour C et al. Front Genet 2019;10:524), sensitivity analyses, or even Bayesian approaches could help formalize such an exploratory analysis.

  • In the presence of repeated exposures with multiple waves of mediators and time-varying covariates, the high dimensionality of the data frequently leads to estimation difficulties due to violation of the positivity assumption. Different strategies have been suggested to improve estimation performance: stochastic counterfactual scenarios, collaborative-TMLE, propensity score truncation optimization, etc (Ju C et al. Stat methods med res 2019;28(6):1741-60). These more recent developments are hardly integrated into currently available software packages. The goal is to be able to combine and implement these methods more easily and compare their performance.

  • The work will be carried out on data available in the “Expanse” European project and on simulated data.
  • Specificities and environment of the job

  • Within the EQUITY team (Biological incorporation, social inequalities, lifecourse epidemiology, cancer and chronic diseases, interventions, methodology) of the CERPOP Unit,
  • Shared office with a computer
  • Access to the CERPOP caclulation server (with R, Stata and SAS software)
  • Benefits

    Duration

    24 months

    Working time

  • Full time
  • 38b0 by week
  • Pay

    From 2 633,57 gross per month depending on professional experience in equivalent positions

    Compensation package

  • 32 days of annual leave and 13 days of RTT (time off)
  • Subsidized collective catering on site
  • Social action and mutual aid committee (social, cultural and sport benefits)
  • Partially reimbursed public transport
  • Selection process

    Send CV and motivation letter to Benoît Lepage (benoit.lepage@univ-tlse3.fr) et Cyrille Delpierre (cyrille.delpierre@inserm.fr)

    Offer Requirements
  • REQUIRED EDUCATION LEVEL
  • Biological sciences: PhD or equivalent

    Skills/Qualifications

    Expertise :

  • Training in biostatistics, or epidemiology with a strong methodological/biostatstical background
  • Be familiar with causal inference approaches and mediation analyses: use of directed acyclic graphs, knowledge of different estimators (g-computation, IPTW, double robust estimation)
  • A general knowledge on different machine learning (data adaptive) alogrithms is useful for the implementation of double robust estimators (TMLE).
  • Theoretical knowledge of asymptotic inference and non-parametric statistical models (asymptotic linearity, influence curves, efficiency theory, empirical processes) can help with the statistical literature describing the derivation of different estimators.
  • Know-how

  • Good knowledge of R Software (how to write functions, modify or program packages if possible)
  • Know how to assess the performance of statistical methods using simulations inspired by real data
  • Know how to prepare a database for statistical analysis
  • Know how to synthesize research results in the form of reports, oral presentations and scientific articles
  • Guarantee the quality and reproducibility of statistical analyses
  • Good level of English, written and oral
  • Good practices in terms of data protection, respect of confidentiality
  • Specific Requirements

    Skills

  • Thorough worker
  • Autonomy, ability to take initiative, adapt and learn
  • Ability to synthesize
  • Ability to work in interaction with other researchers and in a team
  • Desired experience

  • PhD. In addition to experience in mediation analysis, experience with health data and especially longitudinal data would be desirable
  • Degree and training

  • PhD in Biostatistics or Epidemiology
  • Contact Information
  • Organisation/Company: INSERM - DR Occitanie Pyrénées
  • Department: CERPOP - UMR1295
  • Organisation Type: Public Research Institution
  • Website: https:// www. occitanie-pyrenees.inserm.fr/
  • E-Mail: benoit.lepage@univ-tlse3.fr cyrille.delpierre@inserm.fr
  • Country: France
  • City: Toulouse
  • Postal Code: 31000
  • Street: Faculté de médecine, 37 Allées Jules Guesde
  • Similar Jobs

    Add to favorites Read more...

    Contribution To The Design Of The Smartfca-Platform (H/F)

    Research Field: Computer science Mathematics › Algorithms Location: France › VANDOEUVRE LES NANCY FCA is a well-founded mathematical framework based on lattice theory and aimed at data analysis, classification, and knowledge discovery. interested in studying...

    University of Oxford

    United Kingdom Aug 31, 2022

    Add to favorites Read more...

    Bioinformatician

    Big Data Institute, Old Road Campus, Headington, Oxford OX3 7LF We are seeking to appoint a Bioinformatician to join the Big Data Institute, -omic and clinical data, in order to generate predictive models of allograft...

    Inria

    France Aug 30, 2022

    Add to favorites Read more...

    PhD Position F/M Deep Learning Meets Numerical Modelling: AI and Biophysics for Computational Cardiology

    2022-05314 - PhD Position F/M Deep Learning Meets Numerical Modelling: AI The Inria Sophia Antipolis - Méditerranée center counts 34 research teams as The center's staff (about 500 people including The majority of the center's...