Title



     mrmedian -- Weighted median of instrumental variable estimates



Syntax



        mrmedian varname_gd varname_gdse varname_gp varname_gpse [if] [in] [, options]



    options               Description
    ----------------------------------------------------------------------------------------------
      level(#)            set confidence level; default is level(95)
      penweighted         penalized weighted estimator
      reps(#)             number of bootstrap replications to obtain standard error
      seed(#)             seed for random number generator for bootstrapping to obtain standard
                            error
      weighted            weighted estimator



Description



    mrmedian performs unweighted, weighted, and penalized median IV estimator on summary level
    data (i.e. reported genotype-disease [SNP-outcome] and genotype-phenotype [SNP-exposure]
    association estimates and their standard errors for individual genotypes).



    See Bowden et al. (2016) for more information.



    varname_gd is a variable containing the genotype-disease (SNP-outcome) association estimates.



    varname_gdse is a variable containing the genotype-disease (SNP-outcome) association estimate
    standard errors.



    varname_gp is a variable containing the genotype-phenotype (SNP-exposure) association
    estimates.



    varname_gpse is a variable containing the genotype-phenotype (SNP-exposure) association
    estimate standard errors.



Options



    level(#); see [R] estimation options.



    reps(#) specifies the number of bootstrap replications for obtaining the standard error. The
        default is 1000 replications.



    seed(#) specifies the initial value of the random-number seed.  The default is the current
        random-number seed. Specifying seed(#) is the same as typing set seed # before issuing the
        command; see set_seed.



    weighted use weights.



    penweighted use penalized weights.






Examples



    Using the data provided by Do et al. (2013) recreate Bowden et al. (2016), Table 4, LDL-c "All
    genetic variants" median estimates.



    Setup
        . use https://raw.github.com/remlapmot/mrrobust/master/dodata, clear



    Select observations (p-value with exposure < 10^-8)
        . gen byte sel1 = (ldlcp2 < 1e-8)



    Unweighted median estimator
        . mrmedian chdbeta chdse ldlcbeta ldlcse if sel1==1



    Unweighted median estimator with reproducible standard error
        . mrmedian chdbeta chdse ldlcbeta ldlcse if sel1==1, seed(12345)



    Weighted median estimator
        . mrmedian chdbeta chdse ldlcbeta ldlcse if sel1==1, weighted



    Penalized weighted median estimator
        . mrmedian chdbeta chdse ldlcbeta ldlcse if sel1==1, penweighted



Stored results



    mrmedian stores the following in e():



    Scalars        
      e(k)                number of instruments
      e(reps)             number of (bootstrap) replications



    Macros         
      e(cmd)              mrmedian
      e(cmdline)          command as typed



    Matrices       
      e(b)                coefficient vector
      e(V)                variance-covariance matrix of the estimates



References



    Bowden J, Davey Smith G, Haycock PC, Burgess S.  Consistent estimation in Mendelian
        randomization with some invalid instruments using a weighted median estimator. Genetic
        Epidemiology, 2016, 40, 4, 304-314.  DOI



    Do et al. Common variants associated with plasma triglycerides and risk for coronary artery
        disease. Nature Genetics, 2013, 45, 1345-1352.  DOI



Author



INCLUDE help mrrobust-author