100 likes | 344 Views
Analysis of Cross-Sectional Studies. In cross-sectional studies outcome and exposure variables are measured simultaneously, and temporal sequence between cause and outcome cannot be determined.
E N D
Analysis of Cross-Sectional Studies • In cross-sectional studies outcome and exposure variables are measured simultaneously, and temporal sequence between cause and outcome cannot be determined. • In the study described here, diabetic subjects with hypertension were prescribed β-blockers and the investigators wanted to assess the effect on blood levels of HDL-2. • The subjects varied with regard to age, the severity of diabetes, body weight, smoking and alcohol use, and blood levels of triglycerides and C-peptide.
HDL-2 levels (effects of smoking and alcohol use) Mean HDL-2 level is higher among those who use alcohol and amongst non-smokers.
HDL-2 levels and Age With increasing age HDL-2 levels tend to be lower but the relationship is not very strong. R-sq = 3.3 %
HDL-2 levels and Weight With increasing body weight the HDL-2 levels tend to be lower. R-sq.= 4.6%
HDL-2 levels and Beta-blockers Administration of Beta-blocker tends to bring down HDL-2 levels. Bearing in mind that high levels of HDL-2 are protective against heart disease diabetic subjects with hypertension needs careful follow-up
Multiple Regression • The research question is asking the effect of β-blockers on HDL-2 levels. This was shown in the previous slide. But Other factors like life-style, age , body weight and so on also influence HDL-2 levels. Taking all these factors together we get the following as the regression equation. The regression equation is HDL = 0.711 - 0.0824 Beta - 0.0173 Alco - 0.0399 Smok - 0.00455 Age - 0.00214 wt - 0.0444 Trigl. + 0.00463 C-pep - 0.00391 Gluc. S = 0.07745 R-Sq = 59.5% R-Sq(adj) = 54.3%
Regression Diagnostic • Advisable to check that assumptions of regression analysis are not violated. In a clinical observational study one may not be able to do much but at least one knows how much one can depend on the results.
Regression Diagnostic - II • To look for unusual data points one need to plot residuals against fitted values. These two plots are of residuals and standardised residuals against fitted values.