The data here is likely going to be biased towards the people who actually DO check HN all the time even if there are more of the once in a while folks unless you factor in the likelihood each group in seeing this poll before it gets buried.
Figure out how long the poll stays on the home page. For each group, scale that group's number up by the reciprocal of the probability of a person seeing the poll conditional on their membership in that group. (This conditional probability will be equal to min(the amount of time the poll stays on the home page divided by the average length between visits for a member of that group, 1).)
This assumes a person's visits are evenly distributed with time, and that members of all groups are equally likely to respond to the poll if they see it.
Probably less frequent visitors are less likely to have registered accounts and therefore less likely to respond to the poll if they see it.
