Having earned a living as a statistician, we don't like using SQL for
surveys. We have specialized tools that hide the data storage and
give us computations, special missing value rules which are a bitch to
write in CHECK() constraints, etc.
Preference scales with five degrees (very strong, strong, don't care,
weak, very weak) are better than scales with fewer choices; those with
more than 7 degrees are not as repeatable (ask the same question (n)
day later and the profile changes). Nobody who knows what they are
doing would use Aaron's Y/N on a Lemonade survey.
