Distribution of data for respondents’ age and years of education As evident from a visual examination of figure 3, the distribution of ages of respondents is right-skewed. On the other hand, the distribution of years of education for the respondents is left-skewed. Thus, the data for both variables does not conform to the normal distribution. In order to assess whether there are outliers in the data, box-and-whisker plots are used as shown in figure 4 (Morgan 2007: 201). Figure 4.
Foxtrots to assess for distribution of data for age and years of education On visually examining the box-and-whisker plots, the postbox for age affirms that the data is right-skewed due to the long upper whisker while the box plot for years of education has a long lower tail hence left-skewed (Black 2010). Additionally, the postbox for years of education indicates that there are outliers in the data, which might be the possible culprit for keenness. Age Categories and Highest Degree These two variables have been measured at the ordinal level.
In this case, the descriptive analyses on the two variables are restricted to frequencies, mode, median and range. The variable for age categories represents ordered data, at four levels, for the respondents’ age. On the other hand, the variable for highest degree denotes the highest level of education completed by a respondent. The proportions of respondents for each of the levels of the age category and highest degree are as presented in figures 5 and 6. Figure 5. Proportion of respondents for different age categories Almost one thirds of he respondents, 32. 8% (n = 414), are in the age category 45-64. Approximately a quarter of the respondents, 24. 38% (n = 307), are in the age category less than 35, which is almost similar to the proportion of respondents in the age category 35-44 years, 24. 23% (n = 305). The least proportion is made up of respondent in the age category 65+, 18. 51% (n = 233). Figure 6. Proportions of respondents by highest degree obtained On visually examining figure 6, over half of the respondents (52. 18%, n = 657) had attained a higher level while more than one fifths (20. 97%, n = 264) had attained a degree.
Less than 10% had either attained “less than Higher” (9. 85%, n = 124) or a postgraduate (9. 53%, n = 120). The least proportion comprises of respondents who had attained further education (7. 47%, n = 94). Part B: A Model to Predict Attitude to Global Warming In this section, a binary logistic model is fitted on the data in order to predict the continuous and categorical predictors are used to predict a categorical outcome (Carson, 2012, Morgan et al. , 2011) . The categorical outcome is whether a respondent believes that the government should address global warming (O = No, 1 -? Yes).
In running the binary logistic regression, the forward stepwise method (Forward: LARD) is used. In employing the forward method, the initial model consists of only the constant term while the predictors with the most significant score statistics are added into the model (Field, 2009) . This process is continued until score statistics for all the predictors are non-significant. At each of these steps, the variables in the model are evaluated on a set exclusion criterion. In this regard, the likelihood ratio (LARD) criterion has been used, which compares the current model to the model when the predictor is excluded.