The increasing prevalence of obesity in the USA, especially among minority populations, is a serious public health concern. This present study analysed repeated measurements at baseline and at 6 and 12 months on 351 women in the control group and 575 women in the intervention group of the Women's Health Trial: Feasibility Study in Minority Populations. Dynamic random effects models were estimated using the three repeated observations to explain the effects of energy and macronutrient intakes, physical exercise, unhealthy eating habits and socio-economic characteristics on the subjects' body weights and waist and hip circumferences. In both the control and intervention groups, physical exercise was negatively associated with body weight and with waist and hip circumferences, while an index of unhealthy eating habits was positively associated (P<0·05). The proportion of energy derived from carbohydrate and from saturated and monounsaturated fat were often significant predictors of body weight and of waist and hip circumferences in the two groups. The results indicated that nutrition education programmes for improving eating habits and increasing physical exercise can reduce obesity prevalence in the USA.