THIS SITE

You are here

Significance testing

Statistical comparisons between groups/years were tested using regression analysis, with p<0.05 as the cut-off for significance. Bootstrap weights were used to perform significance testing between groups or between the latest two years, where they were available (CTADS 2013 and 2015, CTUMS 2001 to 2012, CSTADS 2014-15, and YSS 2004-05 to 2012-13). For binary response variables, statistical comparisons were performed using the SURVEYLOGISTIC procedure in SAS statistical software (Version 9.4) for CSTADS or the svy: logit command in Stata (Version 14.2) for CTADS. For the continuous variable of cigarettes per day, comparisons were performed using the SURVEYREG procedure in SAS for CSTADS or the svy: regress command in Stata for CTADS. The change in the specification of the bootstrap weights from CTADS 2013 to CTADS 2015 necessitated using the variances (each calculated using their own set of bootstrap weights) to perform a z-test when testing the differences in estimates between CTADS 2013 and CTADS 2015. Comparisons of prevalence rate and cigarettes per day (CPD) over the time period 1999-2015 (#4, #5, #13 and #14) were tested using a dataset of the prevalence rates and CPD by year. The GLM procedure in SAS was used with the yearly prevalence rate regressed on year for #4, the log of yearly prevalence rate regressed on year for #5, yearly CPD regressed on year for #13 and log of yearly CPD regressed on year for #14.

Where statistical testing has been performed, comparisons are marked with a superscript number, which refers to a p-value that can be found in the Index of Statistical Tests. Throughout the report, the term “significant” has been reserved for instances where statistical testing has been performed at the 5% of level of significance (i.e., p<0.05).