Title



     mrratio -- Instrumental variable ratio (Wald) estimator



Syntax



        mrratio #gd #gdse #gp [#gpse #cov] [, options]



    options               Description
    ----------------------------------------------------------------------------------------------
      eform               exp() estimate and CI limits
      fieller             CI using Fieller's Theorem
      level(#)            set confidence level; default is level(95)
      nome                NOME assumption



Description



    mrratio implements the standard instrumental variable ratio (Wald) estimator.



    mrratio can generate confidence interval limits using either a standard error from a Taylor
    series expansion (the default), a standard error using the NOME (no measurement error) in the
    genotype-phenotype (SNP-exposure) association assumption, or using Fieller's Theorem.



    The user needs to have summary level genotype-disease (SNP-outcome) and genotype-phenotype
    (SNP-exposure) association estimates and their standard errors. If assuming NOME the standard
    error of the genotype-phenotype (SNP-exposure) association is not required.



    It is also optional to provide the covariance between the genotype-disease (SNP-outcome) and
    genotype-phenotype (SNP-exposure) estimates. If these are from independent samples then this
    covariance is zero.



    #gd genotype-disease (SNP-outcome) association estimate.



    #gdse standard error of the genotype-disease (SNP-outcome) association estimate.



    #gp genotype-phenotype (SNP-exposure) association estimate.



    #gpse standard error of the genotype-phenotype (SNP-exposure) association estimate.



    #cov covariance between the genotype-disease (SNP-outcome) and the genotype-phenotype
    (SNP-exposure) estimates.



Options



    eform specifies reporting the exponentiated ratio estimate and confidence interval limits.



    fieller specifies deriving the confidence interval using Fieller's Theorem.



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



    nome specifies the NOME (no measurement error in the genotype-phenotype [SNP-exposure]
        association) assumption.



Examples



    Hypothetical example
        . mrratio 1 .5 1 .25



    With NOME assumption
        . mrratio 1 .5 1 .25, nome



    With NOME assumption
        . mrratio 1 .5 1, nome



    Exponentiate estimate and CI limits
        . mrratio 1 .5 1 .25, eform



    CI using Fieller's Theorem
        . mrratio 1 .5 1 .25, fieller



Stored results



    mrratio stores the following in e():



    Scalars        
      e(fiellerres)       Result code from the Fieller CI: . fieller not specified; 1 closed CI; 2
                            union of two intervals; 3 CI is whole real line
      e(level)            level specified in first call to mrratio.



    When fieller specified additional scalars:
      e(lowerci)          lower CI limit
      e(upperci)          upper CI limit



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



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



Author



INCLUDE help mrrobust-author