This article proposes new model-fitting techniques for quantiles of an observed data sequence, including methods for data segmentation and variable selection. The main contribution, however, is in providing a means to perform these two tasks simultaneously. This is achieved by matching the data with the best-fitting piecewise quantile regression model, where the fit is determined by a penalization derived from the minimum description length principle. The resulting optimization problem is solved with the use of genetic algorithms. The proposed, fully automatic procedures are, unlike traditional break point procedures, not based on repeated hypothesis tests, and do not require, unlike most variable selection procedures, the specification of a tuning parameter. Theoretical large-sample properties are derived. Empirical comparisons with existing break point and variable selection methods for quantiles indicate that the new procedures work well in practice.

Background

Obesity and overweight have increased dramatically in the United States over the last decades. The complexity of interrelated causal factors that result in obesity needs to be addressed within the cultural dynamic of sub-populations. In this study, we sought to estimate the effects of a multifaceted, community-based intervention on body mass index (BMI) among Mexican-heritage children.

Methods

Niños Sanos, Familia Sana (Healthy Children, Healthy Family) was a quasi-experimental intervention study designed to reduce the rate of BMI growth among Mexican-heritage children in California's Central Valley. Two rural communities were matched based on demographic and environmental characteristics and were assigned as the intervention or comparison community. The three-year intervention included parent workshops on nutrition and physical activity; school-based nutrition lessons and enhanced physical education program for children; and a monthly voucher for fruits and vegetables. Eligible children were between 3 and 8 years old at baseline. Intent-to-treat analyses were estimated using linear mixed-effect models with random intercepts. We ran a series of models for each gender where predictors were fixed except interactions between age groups and obesity status at baseline with intervention to determine the magnitude of impact on BMI.

Results

At baseline, mean (SD) BMI z-score (zBMI) was 0.97 (0.98) in the intervention group (n = 387) and 0.98 (1.02) in the comparison group (n = 313) (NS). The intervention was significantly associated with log-transformed BMI (β = 0.04 (0.02), P = 0.03) and zBMI (β = 0.25 (0.12), P = 0.04) among boys and log-transformed BMI among obese girls (β = - 0.04 (0.02), P = 0.04). The intervention was significantly and inversely associated with BMI in obese boys and girls across all age groups and normal weight boys in the oldest group (over 6 years) relative to their counterparts in the comparison community.

Conclusions

A community-based, multifaceted intervention was effective at slowing the rate of BMI growth among Mexican-heritage children. Our findings suggest that practitioners should consider strategies that address gender disparities and work with a variety of stakeholders to target childhood obesity.

Trial registration

clinicaltrials.gov Identifier: NCT01900613 . Registered 16th July 2013.