help episens  
-------------------------------------------------------------------------------

Title

episens - Basic sensitivity analysis of epidemiological results

Syntax

Cohort studies and Case-control studies

episens var_case var_exposed [var_time] [if] [in] [weight] [, options]

episensi #a #b #c #d [, options]

Options

options Description ------------------------------------------------------------------------- Options study(cc|cs|ir) specify the type of study unconf(#pexp #punexp #rrcd) analysis of unmeasured confounding miexp(#seca #spca #senc spnc) analysis of misclassification of the exposure corder(#) corrections order format(%fmt) display format for numbers ------------------------------------------------------------------------- Description

episens provides basic sensitivity analysis of the observed relative risk for unmeasured confounding and misclassification of the exposure, or both. episensi is the immediate form of episens; see immed.

Options

+-------+ ----+ Model +------------------------------------------------------------

study(cc|cs|ir) specifies the type of study: cc (case-control data), cs (cumulative-incidence data), or ir (incidence-rate data). The default is cc. See epitab for more information.

unconf(#pexp #punexp #rrcd) specifies the analysis of unmeasured confounding, where #pexp is the prevalence of the unmeasured confounder among the exposed subjects; #punexp is the prevalence of the unmeasured confounder among the unexposed subjects; and #rrcd is the relative risk between the unmeasured confounder and disease.

miexp(#seca #spca #senc spnc) analysis of misclassification of the exposure, where #seca is the sensitivity among the cases; #spca is the specificity among the cases; #senc is the sensitivity among the non-cases; and #spnc is the specificity among the non-cases.

corder(#) specifies the correction order: 1 means first correct for unmeasured confounding and then for misclassification of the exposure; 2 means first correct for misclassification of the exposure and then for unmeasured confounding.

format(%fmt) display format for numbers; default is format(%3.2f).

Examples

// Cumulative incidence data

webuse ugdp.dta cs case exposed [fweight=pop] episensi 30 21 174 184 , st(cs) unc(0.3 0.7 3) episens case exposed [fweight=pop], st(cs) episens case exposed [fweight=pop], st(cs) unc(0.3 0.7 1) episens case exposed [fweight=pop], st(cs) unc(0.3 0.7 3) episens case exposed [fweight=pop], st(cs) unc(0.3 0.7 3) mie(0.9 0.8 > 0.9 0.8)

// Incidence-rate data

sysuse cancer.dta gen exp = drug == 2 ir died exp studytime episens died exp studytime episens died exp studytime , st(ir) unc(0.5 0.2 2) episensi 6 25 209 535 , st(ir)

// Case-control data

episensi 45 94 257 945, st(cc) unc(0.7 0.5 5) episensi 45 94 257 945, st(cc) mie(0.9 0.9 0.8 0.8) episensi 45 94 257 945, st(cc) mie(0.9 0.9 0.8 0.8) episensi 45 94 257 945, st(cc) mie(0.9 0.9 0.8 0.8) episensi 45 94 257 945, st(cc) unc(0.7 0.5 5) mie(0.9 0.9 0.8 0.8) co( > 1) episensi 45 94 257 945, st(cc) unc(0.7 0.5 5) mie(0.9 0.9 0.8 0.8) co( > 2) episensi 45 94 257 945, st(cc) mie(0.8 0.8 0.8 0.8) episensi 45 94 257 945, st(cc) unc(0.7 0.5 5) mie(0.9 0.9 0.8 0.8) episensi 45 94 257 945, st(cc) unc(0.7 0.5 5) mie(1 1 1 1)

References

Greenland S., 1996, Basic methods for sensitivity analysis of biases. Int J Epidemiol, 5(6):1107-16.

Authors

Nicola Orsini, Division of Nutritional Epidemiology, Institute of Environmental Medicine, Karolinska Institutet, Sweden

Rino Bellocco, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Sweden

Sander Greenland, Department of Epidemiology, UCLA School of Public Health, USA

Support

http://nicolaorsini.altervista.org nicola.orsini@ki.se

Also see

Manual: [ST] epitab

Online: epitab, episensrri (if installed)