Question

Our Wheelan readings this week offer an example of a weight-related regression where a very small person could end up having a negative body weight. Likewise, weight ends up being partially determinable by adding variables that have nothing to do with the biology of the individual.

Discuss your explanation for how these factors can be true of our regression analysis, and yet not invalidate the regression itself. Discuss whether or not we should consider these variables as causes of a persons weight, and if not, how should we think about these variables? How do you explain that negative body weight possibility?

Answer 1

Weight in negative seems to be impossible in real life application but let's assume that it is even possible then as well we need to check few things such as ;

If the data belongs to the normal distribution then it's ok because it has both values negative as well as positive but data shouldn't belongs from log-normal or from beta distribution because they doesn't take any negative value.

We should check the homoscedasticity by durbin-watson.

Test for multi-collinearity as well.

But after all this it is possible that a regression variable could take negative value because afterall regression shows the relation between the value of the two variables.