I sent a postal questionnaire to 1618 schools. In total, 220 responded (13.6% response rate). For one question, which had Yes, No and Don't know response options, 217 schools responded (Yes=85, No=127, Don't know=5). I'd like to find the margin of error for Yes, No and Don't know (with 95% confidence) so I can work out what the responses for each would have been if all 1618 schools had responded.
Do I work out the overall margin of error based on the response rate for the whole questionnaire (i.e., 220 of 1618 schools) and apply the upper and lower % to this particular question, or do I use the response rate for each answer in this particular question to find out the margin of error? For example, for Yes, should I use 85/1618, 85/220 or 85/217 to work out the margin of error? There is a big difference in results depending on which one I use.