Discriminant analysis discriminates ethical and unethical behaviour

In order to identify the important variables that predict ethical and unethical behaviour, discriminant analysis was used. Prior research suggested that the variables that affect ethical decisions were attitudes, leadership, the presence or absence of ethical codes of conduct and the organisation’s size. To determine which of these variables were the best predictors of ethical behaviour, 149 firms were surveyed and asked to indicate how their firm operated in 18 different ethical situations. Of these 18 situations, 9 related to marketing activities. These activities included using misleading sales presentations, accepting gifts for preferential treatment and pricing below out-of-pocket expenses. Based on these nine issues, the participating firms were classified into two groups: ‘never practice’ and ‘practice’. An examination of the variables that influenced classification via twogroup discriminant analysis indicated that attitudes and a company’s size were the best predictors of ethical behaviour. Evidently, smaller firms tended to demonstrate more ethical behaviour on marketing issues.

In a study of personal alarms, women were asked to rate eight personal alarms using the following 15 statements:

1 Feels comfortable in the hand.

2 Could be easily kept in the pocket.

3 Would fit easily into a handbag.

4 Could be easily worn on the person.

5 Could be carried to be very handy when needed.

6 Could be set off almost as a reflex action.

7 Would be difficult for an attacker to take it off me.

8 Could keep a very firm grip on it if attacked.

9 An attacker might be frightened that I might attack him with it.

10 Would be difficult for an attacker to switch off.

11 Solidly built.

12 Would be difficult to break.

13 Looks as if it would give off a very loud noise.

14 An attacker might have second thoughts about attacking me if he saw me with it.

15 I would be embarrassed to carry it around with me.

The question was ‘Could these 15 variables be reduced to a smaller number of derived variables, known as factors, in such a way that too much information was not lost?’. Factor analysis enabled these 15 variables to be reduced to four underlying dimensions, or factors that women used to evaluate the alarms. Factor 1 seemed to measure a dimension of size, on a continuum of small to large. Factor 2 tapped into aspects of the appearance of a personal alarm. Factor 3 revealed robustness characteristics, with factor 4 related to hand feel.