Recent social and educational policy debate in the UK has been strongly influenced by studies which have found children’s cognitive developmental trajectories to be significantly affected by the socio-economic status of the households into which they were born. Most notably, using data from the 1970 British cohort study, Feinstein (2003) concluded that children from less advantaged backgrounds who scored high on cognitive tests at 22 months had been overtaken at age 5 by children from more advantaged origins, who had scored lower on the baseline test. However, questions have been raised about the methodological robustness of these studies, particularly the possibility that their key findings are, at least in part, an artefact of regression to the mean. In this paper we apply and assess the Growth Mixture Model (GMM) as an alternative approach for identifying and explaining cognitive developmental trajectories in children. We fit GMMs to simulated data and to data from the Millennium Cohort Study to assess the suitability of GMMs for studying socio-economic gradients in developmental trajectories. Our results show that GMMs are able to recover the data generating mechanism using simulated data, where the conventional approach is subject to regression to the mean. Substantively, our MCS findings provide no support for the contention that more initially able children from disadvantaged backgrounds are ‘over-taken’ in cognitive development by less initially able children from more affluent backgrounds. We do, however, find that cognitive developmental trajectories are related to socio-economic status, such that initial class-based inequalities increase over time.

中文翻译：

---

使用成长混合模型对儿童认知发展中的社会经济梯度进行重新评估

最近的英国社会和教育政策辩论受到研究的强烈影响，这些研究发现，儿童的认知发展轨迹受其出生家庭的社会经济地位的影响很大。最值得注意的是，Feinstein（2003）使用来自1970年英国队列研究的数据得出的结论是，来自优势地位较低背景的儿童在22个月时在认知测验中得分较高，在5岁时被优势地位较高的儿童所取代，而这些人的得分较低。基线测试。但是，关于这些研究的方法学鲁棒性，尤其是其关键发现至少部分是回归均值的假象的可能性，已经引起了疑问。在本文中，我们应用和评估了生长混合模型（GMM），作为识别和解释儿童认知发展轨迹的替代方法。我们将GMM拟合到模拟数据和千年队列研究的数据中，以评估GMM在研究发展轨迹中的社会经济梯度时的适用性。我们的结果表明，GMM能够使用模拟数据恢复数据生成机制，而传统方法则需要回归均值。实质上，我们的MCS研究结果不支持以下观点：来自较富裕背景的较不具备最​​初能力的儿童在认知发展中“超接”了来自较弱背景的较具有本能的儿童。但是，我们确实