I am trying to get the socio-demographic characteristics of the areas (buffers) served by restaurants. Using GIS's Network analysis, I created a buffer zone for different sites across NYC. Each site serves different census tracts (CT) in its buffer zone.

For sex, age, and race, I simply summed it up for a buffer level, but I am having trouble with the poverty variable.

I used the federal poverty guideline and census tracts level median family income to dichotomize all the census tracts in the buffer zone as either above or below the poverty guidelines.

I am doing regression at the site level, therefore I need one poverty indicator for each site. So I decided to use the median function to get a value, but for 2-3 sites I get 0.5 as an indicator which makes it harder to add to the regression since poverty is supposed to be binary variable.

Any advice is helpful?