Reply #22: I think it would be more statistically corect to attribute unknowns based on known ratios.

[OneTenthofOnePercent]

I believe you are probably right based on logic - that "unknown" is likely to be "stranger"... but the fair approach is to divide the unknown group amongst all other categories in ratios represented by those categories. This assumes that, knowing the makeup of a well diversified portion of a population, the unknown portion of population should fall within the same ratios/trends.

Example: If you sample 90% of a population and the population prefers "A" over "B" 2:1... then it can be assumed that 6.67% of the unknown population prefer "A" and 3.33% of the unknown population prefer "B".