Pak.j.stat.oper.res. Vol.X No.4 2014 pp361-368 Log-Normal Model is the Best Fitted Model for Duration from Chest Pain to Coronary Artery Disease Diagnosis: An Outcome of Retrospective Cross Sectional Study Mehwish Hussain Department of Research Dow University of Health Sciences, Karachi, Pakistan mehvish.hussain28@gmail.com Nazeer Khan Department of Research Jinnah Sindh Medical University, Karachi, Pakistan nazeerkhan54@gmail.com Mudassir Uddin Department of Statistics, University of Karachi, Karachi, Pakistan mudassir2000@hotmail.com Abstract Cox proportional hazard model is the widely used technique for studying duration data. However, studies showed parametric modeling yields better estimation; though these models are less used due to their complicated application and interpretation. Alongside, the duration from chronic chest pain to the diagnosis of coronary artery disease has not evaluated in the literature. Therefore, this research investigated application of parametric modeling on the current duration while studying outcome from cross-sectional study. Akaike Information Criteria was used to adjudicate different duration models. Keywords: Chest Pain, coronary artery disease, parametric duration analysis, Akaike Information Criteria, log-normal distribution 1. Introduction Coronary artery disease (CAD) is ranked highest among chronic fatal diseases worldwide (Murray and Lopez, 1996). Chest pain is one of the perceptible symptoms for the disease (Bösner et al, 2010; Hussain and Khan, 2011; Hussain et al, 2011). The statistics lie around studying prevalence, risk factors, treatment and hazards of CAD have been sufficiently published (Torpy et al, 2009; Genders and Hunink, 2012). However, hazardous predictors imperiling the shortening of the duration from chest pain to the diagnosis of CAD have not been studied yet (Hussain and Khan, 2011; Hussain et al, 2011). For duration modeling, non-parametric, semi-parametric and parametric methods are used (Kieding et al, 2002; Hussain and Khan, 2011; Hussain et al, 2011). Amongst them, parametric modeling yields good estimate with better interpretation (Morkveld, 2007). Nevertheless, these models are rarely considered to be computed due to two reasons, one is that the distribution of duration must be known and the other reason is that interpretation of estimated coefficient is not much comprehensible (Moghimi- Dehkordi et al, 2008). This study filled the gap while estimating parametric models for the chronic chest pain in patients who later diagnosed with coronary artery disease.