The first slope was used to model the initial survey participation effect (the marked decline after the first wave that reflects learning and increased awareness of health effects of smoking) and http://www.selleckchem.com/products/Dasatinib.html the second slope to model the subsequent linear decline (constant slope, starting from baseline) between assessment Waves 2 and 5. For each of the alternate models, the factor loadings of the level factor to each time-based CPD measure were fixed at 1, while those of the slope factor were fixed to represent the expected pattern of change over the study period as follows: 0, 0, 0, 0, 0 (no growth); 0, 1, 2, 3, 4 (linear); 0, 1, 4, 9, 16 (quadratic); slope 1: 0, 1, 1, 1, 1 and slope 2: 0, 0, 1, 2, 3 (piecewise).

To allow us to model the survey participation effect of the different cohorts, data from the different cohorts were realigned based on individual��s wave of assessment (rather than based on the actual survey years) before they were used in LGC modelling. This data realignment (see Figure 3) would also allow us to model the pattern of change in cigarette consumption beyond the first two assessments without being confounded by the survey participation effect of subsequent cohorts, something that was problematic when modelling on data based on survey years. Detailed descriptions and examples of LGC modelling using structural equation modelling programs are available elsewhere (Duncan, Duncan, Strycker, Li, & Alpert, 1999; McArdle & Bell, 2000; Stoolmiller, 1995). Figure 1. Actual and modelled mean trajectory of daily cigarette consumption by country. Figure 2.

Latent growth curve model of reported cigarettes per day among continuing adult smokers. Note: CPD1-5, square root�Ctransformed cigarette per day for assessment Waves 1�C5; E1�CE5, measurement errors at each assessment wave; d_INT, … Figure 3. Actual mean trajectory of daily cigarette consumption by cohort (top panel) and actual and modelled mean trajectory of daily cigarette consumption overall (bottom panel). Correlates of baseline levels and rate of change of reported CPD were examined by regressing these parameters onto a set of covariates, such as age, sex, country, quit attempts between waves, reported baseline smoking restrictions at both home and workplace, and wave of recruitment.

For the purpose of analyses and to increase sensitivity, reported quit attempts between Brefeldin_A assessment Waves 1 and 2 were used to predict the survey participation effect, while a new variable was derived using the follow-up quit attempt questions to indicate whether the respondents had made at least one quit attempt over the period between assessment Waves 2 and 5. This variable was coded as 1 = Yes and 0 = No to determine its influence on the rate of change in CPD between assessment Waves 2 and 5. Evaluation of Model Fit The adequacy of model fit was assessed by the chi-square statistic or discrepancy function.