International Biometric Society International Biometric Conference, Dublin, 13 – 18 July 2008 A LONGITUDINAL STUDY OF THE EFFECTS OF LIFE-DOMAIN SATISFACTIONS ON LOSS OF SLEEP H. Zahed and M. Ganjali Department of Statistics, Faculty of Mathematical Sciences, Shahid Beheshti University, Tehran, Iran. s.h.zahed@gmail.com and m-ganjali@sbu.ac.ir . Introduction: Longitudinal Studies (in contrast to Cross-Sectional Studies) are studies in which members of the sample are measured repeatedly through time. Hence, the information about the past can be used -alongside the information about the present- to predict the future. Longitudinal Studies also allow the researcher to distinguish what usually is referred to as cohort effect and age effect. Diggle et al (2002) give a thorough review of longitudinal studies, their advantages and common approaches to modelling such data. In this paper a longitudinal approach is taken to study the effects of four aspects of life-domain satisfaction on Insomnia. Sleep problems and Subjective Well-being: Sleep problems also referred to as Insomnia are quite common in modern societies. Some reports suggest that as much as 95% of Americans suffer from Insomnia at some point in their lives. In a 1995 Poll by Gallup, 12% of Americans reported having difficulty sleeping on a frequent basis. The most frequent response to the cause of this loss of sleep was stated as stress/worry. (Gallup Organization, 1995) Reasons for sleep problems can be generally divided into 3 categories: Physical reasons (ageing, pain, use of particular drugs, etc.), Mental and Psychological reasons (worry, stress and depression) and Surroundings (noise, bedroom temperature, uncomfortable bed, etc.). Whilst Physical reasons might need drug-therapy, in case of mental and psychological reasons drug-therapy is usually suggested with precaution. Subjective well-being as a growing area of Psychology focuses on how people evaluate their own lives. There is general agreement in the literature that subjective well-being consists of three components, namely: Positive Affective Appraisal, Negative Affective Appraisal and Life Satisfaction. Life satisfaction is distinguished from affective appraisal in that it is more cognitively driven than emotionally. Researchers generally distinguish between life-domain satisfaction and global life satisfaction. Life-domain satisfaction refers to satisfaction with a specific area of an individual’s life, such as work, marriage etc., whereas judgements of global life satisfaction are much more broad, consisting of an individual’s comprehensive judgement of his/her life as a whole. (Diener et al, 1985) Data: The data used in this paper is extracted from waves 13 (2003), 14 (2004) and 15 (2005) of the British Household Panel Survey (BHPS), which is an annual longitudinal survey of a nationally representative sample of adult (16+) Britons, being carried out since 1991. This survey covers a broad range of social and economical issues, including detailed data on personal health conditions and subjective well-being. From the 13645 individuals present in all three waves, 10923 were included in the analysis due to the elimination of proxy respondents (916 people or 7%) and missing values (1806 people or 13%) which may be ignored. Loss of sleep due to worry was measured by directly asking the question: “Have you recently lost much sleep over worry?”; and the answers recorded in 3 ordered categories: “Not at all”, “No more than usual” and “More than usual”. Satisfaction with four domains of one’s life, namely: Health, Spouse/Partner, Job and Social Life were measured again by ordinal variables with three categories: “Not satisfied”, “Not satisfied/dissatisfied” and “Satisfied”. Other covariates included in the analysis were: Sex, Age, Marital status (Married/Living as a couple/Widowed, divorced or separated/Never married) and Highest educational qualification (Higher or first degree/Other higher degree/Other qualification or no qualification). Original response categories were merged in a logical manner for convenience. Table 1 shows the number of transitions for the response variable “Loss of sleep due to worry” from 2003 to 2004 and from 2004 to 2005. Figure 1 shows these transitions for a sample of size 30 from the data.