740 likes | 932 Views
Avoiding Bias Due to Unmeasured Covariates. Presentations in this series Overview and Randomization Self-matching Proxies Intermediates Instruments Equipoise. Alec Walker. X. T. D. X. Randomization. T. D. X. Randomization. Self-matching. T. D. X. Randomization.
E N D
Avoiding Bias Due toUnmeasured Covariates Presentations in this series Overview and Randomization Self-matching Proxies Intermediates Instruments Equipoise Alec Walker
X T D
X Randomization T D
X Randomization Self-matching T D
X Randomization Self-matching T D Proxies Proxies
A textbook definition from econometrics.
Let O be an outcome (either T treatment or D disease) P be a proxy X be an unmeasured covariate P is a proxy for X with respect to O if thedistribution of O given Pis identical to the distribution of O given P and X A textbook definition from econometrics. • Which is to say that X adds no information about O, if you know P.
Let O be an outcome (either T treatment or D disease) P be a proxy X be an unmeasured covariate P is a proxy for X with respect to O if thedistribution of O given Pis identical to the distribution of O given P and X A textbook definition from econometrics. • Which is to say that X adds no information about O, if you know P. Note that O, PandX could all be multidimensional, that is vectors of outcomes, proxies and unmeasured covariates, respectively. This definition could also be conditioned on other, measured covariates.
Proxy variables are Correlates of an unmeasured covariate That are useful to the extent that they capture the influence of the unmeasured covariate on a third characteristic Control for a proxy replaces control for the unmeasured covariate
Interview responses may be proxies for Historical measurements (diet, smoking, alcohol …) Internal states Genetic traits Biological markers are proxies for biological processes Age, sex, SES are stand-ins for their many correlates . Examples of proxies
Interview responses may be proxies for Historical measurements (diet, smoking, alcohol …) Internal states Genetic traits Biological markers are proxies for biological processes Age, sex, SES are stand-ins for their many correlates . Examples of proxies In diabetics, retinal vascular disease is a proxy for vascular disease more generally and is easily ascertained by funduscopic examination.In looking at determinants of myocardial infarction, control for retinal vascular disease could represent control for coexisting vascular pathology.
Source: US Department of Veterans Affairs Early diabetic retinopathy https://www.myhealth.va.gov/mhv-portal-web/anonymous.portal?_nfpb=true&_pageLabel=commonConditions&contentPage=va_health_library/diabetic_retinopathy_advanced_info.html
Source: US Department of Veterans Affairs Early diabetic retinopathy microaneurysms https://www.myhealth.va.gov/mhv-portal-web/anonymous.portal?_nfpb=true&_pageLabel=commonConditions&contentPage=va_health_library/diabetic_retinopathy_advanced_info.html
Source: US Department of Veterans Affairs Advanced diabetic retinopathy https://www.myhealth.va.gov/mhv-portal-web/anonymous.portal?_nfpb=true&_pageLabel=commonConditions&contentPage=va_health_library/diabetic_retinopathy_advanced_info.html
X D T
X P D T
X X P T D D
X X P UD T D D UT
X X P UD T D D UT
UX X P UD T D UT
UX X X P UD T D D UT
UX X X P UD T D D UT
(Unmeasured) Severity of Diabetes Coronary artery disease Retinal vascular disease UD Acute myocardial infarction Thialozinediones for diabetes UT
Without mechanistic information, for each of these situations, ( covariate causes proxyproxy causes covariateboth caused by a third factor ) … the proxy looks like a transformation of the predictor, with added error. Proxy value = f(Predictor value) + error
An accurate proxy Treated Untreated The true value of the unmeasured covariate is a predictor of treatment
An accurate proxy The proxy predicts treatment almost as well as does the true value. Treated Untreated The true value of the unmeasured covariate is a predictor of treatment
An accurate proxy The proxy almost perfectly represents the value of the unmeasured covariate. Treated Untreated
An accurate proxy Treated Untreated
An accurate proxy Treated Untreated
An accurate proxy The proportion of treated among subjects in a particular small range of proxy values
An accurate proxy The proportion of treated among subjects in a particular small range of proxy values
An accurate proxy The proportion of treated among subjects in a particular small range of proxy values … is the same as the proportion of treated among subjects in the corresponding small range of true values.
An accurate proxy The true value does not provide further information, if you know the proxy.
Two accurate proxies Treated Untreated
Two accurate proxies Treated Untreated Two good proxies are highly correlated with one another.
Two accurate proxies Either proxy provides good prediction of treatment. Treated Untreated
Proxies with substantial random error Treated Untreated Untreated
Proxies with substantial random error Treated The proxy is still correlated with the unknown measure. Untreated Untreated
Treatment is still associated with higher values of the proxy, but thediscriminationis muchworse. Proxies with substantial random error
Proxies with substantial random error Treated Untreated
Proxies with substantial random error Treated The correlation between the two proxy measures is still evident. Untreated
Both proxies show poor discrimination between treated and untreated. Proxies with substantial random error Treated Untreated
Proxies with substantial random error The two proxies can be combined into a function that discriminates better than either proxy alone. Treated Untreated
A textbook definition from econometrics.
Let O be an outcome (either T treatment or D disease) P be a proxy X be an unmeasured covariate P is a proxy for X with respect to O if thedistribution of O given Pis identical to the distribution of O given P and X. A textbook definition from econometrics.