The BEARS.MTW dataset comes with MINITAB, with the data having been provided by Gary Alt. The objective is to develop a way for hunters to estimate the weight of a bear since a bear cannot be weighed in the wilderness, but other measurements could be taken if the bear was dead or at least incapacitated. Determine if a useful regression model can be developed for this purpose, paying particular attention to the following.
(a) You will note that there are many missing observations on the variable AGE. When data are missing, it is important to determine whether data are missing at random or missing systematically. Of course, this determination generally cannot be easily made without imputing missing data. In this instance, would you be inclined to use a variable that is highly correlated with age, if there is one, instead of age, so as to be able to use more observations in building the model?
(b) You will notice that the weight of the 143 bears varies greatly. Does your model apply equally well across the wide range of weights or would you suggest that different models be used for different weight intervals? If so, how would you proceed?