We categorised pharmacokinetic parameters as 'wild type' (*1/*1) and 'variant type' (single and double variant) genotypes, consistent with pharmacogenetic studies on nicotine, another substrate of CYP2A6 (Schoedel et al, 2004; Nakajima et al, 2006; Mwenifumbo et al, 2007).
Haematological parameters categorised as (abnormal vs normal) were compared by HIV status and among HIV+ women by CD4 count category using proportions.
Individual practices were categorised by several parameters: Practices governed by Māori organisations were labelled Māori provider practices and were compared to other (not Māori provider) provider practices.
The parameters are categorised into three criteria groups, namely physical, environmental and economic criteria and then the weights are verified by a group discussion with the experts for final weight consensus making them more objective.
When the QUS parameters were categorised into quintiles, compared with those in the highest quintile of BUA or SOS, the risk of death was highest in those in the lowest quintiles.
The model selection process is categorised into: Kernel selection; Parameter grid search; Feature selection.
The improvement potential of those parameters can be categorised as 'low', 'moderate' and 'high'high
One-way ANOVA with post-hoc Scheffe tests were used to compare DTI parameters between groups categorised as normal/mildly, moderately or severely abnormal WM.
To select optimal cutoff values for each of the FISH parameters, patients were categorised into FISH positive and FISH negative groups based on each of the possible cutoff values.
Hazard models were also used to assess the association between the QUS parameters (per SD, categorised into quintiles and dichotomised as lowest quintile versus higher quintile) and mortality (all-cause and CVD) initially unadjusted, then serially adjusting for age and centre, then factors found to be associated with mortality in the earlier analysis.
