Survival data are often clustered into groups, such as couples, families, communities, and geographical regions. Observations from same cluster usually share certain unobserved characteristics and as a result tend to be correlated. In analysis if associations among observations are ignored, standard error of the estimates of parameters of interest may be incorrect. In this study, Cox’s proportional hazards model and frailty models are used for the analysing Bangladesh Demographic and Health Survey (BDHS, 2004) data. In frailty models, correlations among observations are considered. In this study child mortality data are analyzed for all the divisions of Bangladesh separately. The results indicate that, in Barisal division child mortality rate is higher compared to the other divisions. The results also indicate that, the mortality is significantly higher in rural areas compared to the urban areas of Sylhet division. For Chittagong division mother age at birth is found to be significant for child mortality rate. For Sylhet and Khulna division child mortality rate is found to be lower for literate mothers compared to illeterate mothers.