Daniel Wright, I suspect that they seem to be undergraduate homework questions, but they are not really homework. During their training people might not have understood what they were supposed to learn and do. It seems like 'Everything You Always Needed to Know About Statistics (But Were Afraid to Ask)'. I do not mind to help, but in those cases I will recommend a book or internet page. A college was giving a statistics course and had 4 hours to teach stuff that took me several books and I still did not get it all. He gave the summary of the summary of the summary and for those students most of the questions that look like undergraduate homework remain a mystery until the are required to do it themselves.

Sometimes it is a little more complicated as with measurement levels and type of analysis one can, must or could do. The book I want to read is Krantz et al Foundations of Measurement Vol 1, but I just can't get rid of my education as it tries to explain to me that 1+1= 2 only under specific conditions. Since primary school I have lived in a world were 1+1=2 and I can not shake it off. Operations that are allowed for interval data are also allowed for ordinal data as these will not change the order of the data. But not all transformation that are allowed for ordinal data are allowed for interval data.

Is it always necessary to perform a logistic regression on binary data? The answer is no, it is better and more appropriate, but not always impossible. Interpretation might be more problematic when you do a normal regression analysis. A leading expert told me that when the predicted values from a model are outside the range of permissible values, the analysis is not correct. He told it, but one can not read it in (his) textbooks. So if the predicted value is 1 in a normal regression analysis, you have performed the wrong analysis.