Premium
The application of subset correspondence analysis to address the problem of missing data in a study on asthma severity in childhood
Author(s) -
Hendry G.,
North D.,
Zewotir T.,
Naidoo R.N.
Publication year - 2014
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.6189
Subject(s) - asthma , epidemiology , multiple correspondence analysis , medicine , correspondence analysis , tobacco smoke , missing data , set (abstract data type) , environmental health , association (psychology) , demography , pediatrics , psychology , statistics , computer science , pathology , mathematics , sociology , psychotherapist , programming language
Non‐response in cross‐sectional data is not uncommon and requires careful handling during the analysis stage so as not to bias results. In this paper, we illustrate how subset correspondence analysis can be applied in order to manage the non‐response while at the same time retaining all observed data. This variant of correspondence analysis was applied to a set of epidemiological data in which relationships between numerous environmental, genetic, behavioural and socio‐economic factors and their association with asthma severity in children were explored. The application of subset correspondence analysis revealed interesting associations between the measured variables that otherwise may not have been exposed. Many of the associations found confirm established theories found in literature regarding factors that exacerbate childhood asthma. Moderate to severe asthma was found to be associated with needing neonatal care, male children, 8‐ to 9‐year olds, exposure to tobacco smoke in vehicles and living in areas that suffer from extreme air pollution. Associations were found between mild persistent asthma and low birthweight, and being exposed to smoke in the home and living in a home with up to four people. The classification of probable asthma was associated with a group of variables that indicate low socio‐economic status. Copyright © 2014 John Wiley & Sons, Ltd.