
DEXSAGE

DEXSAGE: Project uses complex statistical techniques to analyse older adults’ daily experiences
Most commonly datasets collected using experience sampling method (ESM) have been analysed using ANOVA and related techniques. This project has employed complex statistical techniques including multilevel mixed and fixed effects models and latent class analysis using generalized structural equation modelling (gsem) in Stata 16 to investigate different aspects of older adults’ daily behaviours and their subjective meaning.
Experience sampling method has been used to obtain the in-situ assessment of particular experiences or emotional states an individual experiences throughout the day. The relatively complex structure of the data generated via this data collection technique yields some analytical challenges. In ESM observations are not independent but they are nested within individuals. For this reason, standard analytical techniques such as regression analysis cannot be employed. This opens up opportunities for more sophisticated statistical techniques to be used.
The present study makes full use of the data and accounts for its complexity. Three papers were produced, each of which has accounted for the multilevel structure of the data in a different way:
1/ Multilevel mixed effects models were used to analyse the episode-level determinants of activity enjoyment;
2/ Generalized structural equation models were used to analyse individual-level behaviours using a combination of count and binomial variables derived from repeated-measures data;
3/ Multilevel logistic regression models were used to analyse which activities were more likely to be reported as best, worst or main activities compared to being reported as random activity
Finally, similar to techniques used in time use data analyses, simple individual-level measures were generated to account for broader characteristics of each day, including individual-level variance in stress, enjoyment or perceived activity importance
This study is thus innovative also in terms of the data analysis techniques.
References
Verhagen, S.J., Hasmi, L., Drukker, M., van Os, J. and Delespaul, P.A., 2016. Use of the experience sampling method in the context of clinical trials. Evidence-based mental health, 19(3), pp.86-89.
Keywords
Experience sampling method, multilevel mixed effects modelling, generalized structural equation modelling