Methodology

View the most recent version.

The multivariate model

The model's independent variables included: age; sex; marital status; presence of children under the age of 18 in the household; education level; geographic location; duration of use (measured in number of years online); frequency of use; intensity (measured in hours online per week) and breadth of use (measured by number of reported Internet activities). To avoid multi-collinearity in the model, income was not included due to a high correlation with education.

The independent variables, followed by the reference groups in parentheses, are:

• Age (less than 45 years)
• Sex (Female)
• Marital Status (Not married)
• Presence of children under the age of 18 in the household (No children)
• Education level (High school or less)
• Geographic location (Rural)
• Duration of Internet use (Online less than 5 years)
• Frequency of Internet use (Online less than once a day)
• Intensity of Internet use (Online less than 5 hours per week)
• Breadth of Internet use (Between 1 and 9 Internet activities)

Each of the variables was entered into the model in one step to determine their unique effect, while the effects of the others were being held constant. The final model was found to be significantly better than the 'null' model (intercept only), with most of the variables exhibiting predictive ability for the behaviour in question, namely, the use of the Internet to access GOL information.

Interpreting odds ratios

The purpose of logistic regression is to estimate the probability of an event occurring (for example, accessing government information online) based on a set of explanatory variables. This technique allows the relationship between each explanatory variable and the event to be examined, while holding all other specified variables constant. Differences between population groups are expressed in terms of odds ratios.

The odds of an event are defined as the ratio of the probability that an event occurs, to the probability that it fails to occur. In this case, the event being considered is "use of the Internet to access government online information". Odds ratios indicate whether certain variables increase or decrease the odds of using the Internet to access government information, compared to a reference group, controlling for all other explanatory variables in the model (see Chart 5).

An odds ratio of 1 represents equal odds for the comparison groups of engaging in this activity. Odds ratios with values below 1 indicate less chance of accessing government online information, and odds ratios larger than 1 represent an increased chance. This article used bootstrap weights to estimate the standard errors to account for the complex sample design used in the survey. The "don't know/refused" responses were excluded from the analysis.