Close  
Figure 2 :Comparison between actual and predicted CAC score distributions among a subset of the study population using three different modeling strategies. Actual prevalence measurements were from the 58 to 62 yearold nonsmoking women in our study sample with hypertension, high cholesterol level, and no diabetes (n = 127). The "twostage model predictions" use the coefficients presented in Tables 2 and 3 (the full model). The Ln(CAC+1) model predictions are from a linear regression model including all conventional CHD risk factors using Ln(CAC score +1) as a continuous outcome in a onestep modeling process (coefficients not presented). The Tobit model uses the cuberoot of the CAC score as a continuous outcome for linear regression analysis, but assumes that scores at or below zero have been censored (coefficients not presented). Pvalues refer to a X2 test with 3 degrees of freedom comparing the expected frequencies based on each model with the observed frequencies. Lower pvalues indicate a poorer model fit. CAC –Coronary artery calcium; CHD – Coronary heart disease; Ln – Natural logarithm. 
