 |
 |
A reminder of the details of the next Biostatistics group seminar at The University of Manchester being held tomorrow are below. All are welcome to attend this free seminar and there is no need to register. Date: Wednesday 16th March 2011 Time: 2.00pm-5.00pm (tea/coffee will be served mid-afternoon) Venue: Manchester Dental Education Centre (MANDEC), Higher Cambridge Street, Manchester Topic: Latent variable modelling Programme: 14.00 - 14.50 Dr Bianca De Stavola, London School of Hygiene and Tropical Medicine "Intergenerational correlations in size at birth and the contribution of environmental factors: the Uppsala Birth Cohort Multigenerational Study, Sweden, 1915-2002" Size at birth in parents and their children are known to be correlated, reflecting in part the influence of foetal and maternal genes. Socio-demographic factors, regarded as aspects of shared environment across generations, would also be expected to contribute, but evidence is limited. The aim of this study is to quantify the role of shared environment in explaining intergenerational correlations in birth weight and length, using data across three consecutive generations from the Uppsala Birth Cohort Multigenerational Study. This includes birth and socio-demographic data on 7,657 singletons, born in Uppsala 1915-1929 (G1), and their grandchildren (G3). Standard regression and biometrical models were used to study the correlations in size at birth of G1-G3 pairs. The data showed stronger correlations in maternal than paternal pairs for birth weight but not for birth length. These correlations were not reduced by adjustment for socio-demographic factors in regression models. In contrast, significant shared environment contributions to the intergenerational correlations were identified in biometrical models, averaging 14% for both birth measures. These models are a particular specification of structural equation models where latent variables are specified to capture the genetic and environmental factors shared by relatives. The results show that the two approaches lead to partly inconsistent results. Reasons of why this happens will be discussed and some general conclusions will be drawn. This is work in collaboration with Ilona Koupil (Karolinska Institute and Stockholm University) and David Leon (LSHTM). 14.50 - 15.40 Dr Tim Croudace, University of Cambridge "Psychometric epidemiology, Rasch psychiatry and advances towards well-being" Psychometric measurement models from educational testing have diversified over the years into health, quality of life and psychiatric research. Modern software developments make it possible for researchers to combine many useful model components in novel and flexible ways; the universe of models is large. However several distinct model types remain somewhat software bound in terms of application areas and communities. Recent experiences at the forefront of software capabilities and from interacting with UK psychometric communities will be summarised in the context of measurement challenges in mental health. Latent variable models discussed include latent class, trait and hybrid models. Applications relate to psychological distress and positive mental health and the notion of categorical versus dimensional spectrum from Rasch and general latent variable modelling perspective. 15.40 - 16.00 Tea/coffee break 16.00 - 16.50 Prof Andrew Pickles, Institute of Psychiatry, KCL "Interactions, effect moderation and random coefficients in latent variable models: identifying patterns of prenatal vulnerability" Following the foetal origins hypothesis of Barker there is now wide & intense interest in developmental programming that can have both adaptive benefits and costs far into the organism's future. One of these programming mechanisms involves the HPA system, where it has been found that in rodents and other animals maternal behaviour can modify genetic predisposition to anxiety through methylation of the DNA - an epigenetic process. Other possibilities include the process of brain-stem maturation setting levels of autonomic response to stress (vagal tone and vagal withdrawal) which might also be modified by post-natal parent behaviour. We explore this possibility in humans using the Wirral Child Health and Development Study and the GLLAMM analysis framework. We estimate a model in which a maternal behaviour latent variable derived from 4 ordinal items by an Item-Response Theory measurement model, moderates the impact of both maternal depression at 32-weeks of pregnancy and poor intra-uterine growth, on 6-month infant vagal tone and on vagal withdrawal in response to a social stressor. GLLAMM - Generalized Linear Latent and Mixed Model ------------------------ You may leave the list at any time by sending the command SIGNOFF allstat to [log in to unmask], leaving the subject line blank.