How many significant figures should there be when performing calculations using discrete numbers (e.g. counts)? I understand significant figures, but I can't find information on this specific case anywhere.
For example, if we surveyed 985 people and found that 9 of them disliked ice cream, how many sig figs would we include when representing that as a proportion or a percentage? Would it be one sig fig (0.009 or 0.9%), or could it potentially be many more since we know that we have 9.000000... people?
Note:
I don't think it should default to three sig figs (for the total number of people surveyed), because this could also be looked at in terms of people who like ice cream, which would be 976 out of 985. This results in 0.991 or 99.1%, which gives the same level of uncertainty ($1 - 0.991 = 0.009$). So, defaulting to the total number of people for sig figs would give less uncertainty simply because we chose to look at the proportion disliking ice cream rather than the proportion liking—which proportion we focus on doesn't change the level of uncertainty we have about the number.
