**************** **************** ** tabodds ** Tabulate Odds **************** **************** Overview: The "tabodds" command is used to obtain summaries and inference for case-control or cross-sectional designs. It tabulates the odds of disease for the levels of a categorical exposure (or explanatory) variable. This is particularly useful for looking at the association between disease and a multi-category exposure variable. The "or" option is useful for computing odds ratios that compare each level >1 of the exposure variable to the first level, or reference level (ie. odds ratio comparing Evar=j to Evar=1 for j=2,3,...). This procedure can be used to obtain two useful tests: (1) A test of homogeneity of odds -- this is a variant on the standard Pearson's chi-square test. (2) A trend test for a linear trend of the log odds against the levels of the explanatory variable. Usage: "tabodds Dvar Evar" Where - Dvar is the disease variable (1=disease, 0=control) Evar is the exposure variable (multiple levels possible) Summaries: The "tabodds" command returns an estimate of the disease odds, p/(1-p), for each level of the exposure variable. A confidence interval is also calculated. A test of homogeneity, and a test for trend are reported. Using the "or" option will compute an odds ratio comparing level=j to level=1 (see below). Options: (1) "tabodds Dvar Evar [fweight=freq]" -- this is used when the data are in a "grouped" format with the number of cases/controls that have a certain covariate combination given by the variable "freq". See notes pg 52 for cc example. (2) "tabodds Dvar Evar, or" -- this option leads to odds ratio summaries. The reference group is the lowest level of the Evar, and the odds ratios are: odds(j)/odds(1) where each category, level=j, is compared to the reference category, level=1. The reference level can be controlled using "base" (see next). (3) "tabodds Dvar Evar, or base(#)" -- the base option allows the reference category to be changed. For example, "base(4)" would use Evar=4 as the reference group, and odds ratios would report: odds(j)/odds(4). **************** **************** ** mhodds ** Mantel-Haenszel Methods **************** **************** Overview: The "mhodds" command is used to obtain summaries and inference for case-control or cross-sectional designs. It calculates an estimate of the odds ratio (disease & exposure) for each level of a stratifying variable. A test for the common odds ratio, and an estimate of the common odds ratio are given. In addition, a test of homogeneity of the odds ratios is computed. One key advantage of "mhodds" is the ability to control for multiple stratifying variables. Usage: "mhodds Dvar Evar, by(Svar)" Where - Dvar is the disease variable (1=disease, 0=control) Evar is the exposure variable (1=exposed, 0=unexposed) Svar is the stratifying variable (multiple levels) Summaries: The "mhodds" command returns the stratum specific odds ratio, a test for each stratum specific odds ratio, a confidence interval for each stratum specific odds ratio, and an adjusted, or common odds ratio estimate, test, and confidence interval. Options: (1) "mhodds Dvar Evar [fweight=freq], by(Svar)" -- this is used when the data are in a "grouped" format with the number of cases/controls that have a certain covariate combination given by the variable "freq". See notes pg 52 for cc example. (2) The overall adjusted estimate can be obtained directly by not using the "by()" option: "mhodds Dvar Evar Svar" This returns the summary odds ratio and tests without showing the stratum-specific odds ratios. No homogeneity test is provided. Similarly, we can adjust for multiple variables: "mhodds Dvar Evar Svar1 Svar2" -- this adjusts for both Svar1 and Svar2 when computing the adjusted OR. No test of homogeneity is given. "mhodds Dvar Evar Svar1, by(Svar2)" -- this computes an a disease odds ratio that is adjusted for Svar1 within each level of Svar2, and then combines these estimates to form a single common odds ratio. A homogeneity test is calculated for the adjusted odds ratios across the levels of Svar2. **************** **************** ** cc ** Using "cc" for Mantel-Haenszel Methods **************** **************** Overview: The "cc" command is used to obtain summaries and inference for case-control or cross-sectional designs. Using the "by(Svar)" option allows Mantel-Haenszel estimates to be obtained. Usage: "cc Dvar Evar, by(Svar)" Where - Dvar is the disease variable (1=disease, 0=control) Evar is the exposure variable (1=exposed, 0=unexposed) Svar is the stratifying variable (multiple levels) Summaries: The "cc" command returns the stratum specific odds ratio and confidence interval, the Mantel-Haenszel weights, the crude odds ratio and the adjusted odds ratio. In addition, both a test of homogeneity of odds ratios, and the Mantel-Haenszel test for the common odds ratio are computed. Options: (1) "cc Dvar Evar [fweight=freq], by(Svar)" -- this is used when the data are in a "grouped" format with the number of cases/controls that have a certain covariate combination given by the variable "freq". See notes pg 52 for cc example. (2) "cc Dvar Evar, by(Svar) bd" -- the "bd" option allows calculation of the Breslow-Day test of homogeneity of the odds ratio. This test is more appropriate than the Mantel-Haenszel test when the stratum specific 2x2 Tables are sparse.