Many women who are initially thought to have angina turn out to have normal coronary angiograms, that is they are found not to have angina after all. A study was carried out to assess the feasibility of a preliminary screening test. For a large number of patients who were thought to have angina, information on a number of possible risk factors was collected and then their subsequent angina status was recorded. The data is available as an R data frame entitled angina and contains the following information: status: whether woman turns out to have angina (yes/no) age: age of a ?woman smoke: smoking status (1=current-, 2=ex-, 3=non-smoker) cig: current average number of cigarettes per day hyper: hypertension (1=absent, 2=mild, 3=moderate) angfam: family history of angina (yes/no) myofam: family history of myocardial infarction (yes/no) strokefam: family history of stroke (yes/no) diabetes: does woman have diabetes? (yes/no) Missing values are coded as NA. The main aim of this study was to try to find out which, if any, of the health variables, are associated with angina and whether some subset of them could be used to help predict the dependent variable angina status. The accompanying document on the `Model selection through backward elimination’ is going to be useful for that purpose. More specifically, it would be helpful to be able to estimate the risk/probability that a woman with a particular combination of these health variables truly has angina. If such a scheme of estimating risks can be constructed, is it likely to be useful? i.e. is it good at predicting whether a woman has angina or not (since the treatment of angina is expensive)? In addition, it would be of interest to estimate the individual effects of important variables. For example, if smoking seems to be a risk factor, then what is the odds of a smoker having angina relative to a non-smoker? What about ex-smokers and light smokers?