The Belfast Telegraph's random, non-weighted sample of opinion at the Sinn Fein conference was a survey, not a poll.
Sinn Fein estimated that 2,000 party members attended the ard fheis. On this reckoning, 50 would represent a 2.5% sample.
In Northern Ireland, LucidTalk opinion poll-projects target a sample of 1,250 out of an adult population of about 1,200,000, i.e. about 0.1%.
Polling science has shown this to be mathematically representative, and similar or smaller sample sizes are used elsewhere, such as in the UK or US.
However, the target samples for our Northern Ireland-wide polls are carefully balanced by religion, area, age groups, and other demographic criteria. This ensures that the error is only 3-3.6% in terms of being representative of the total population.
With a random sample of 50 at the Sinn Fein conference the margin of error could be as high as 8-9.5% of any given score. This means that if 30% of those surveyed gave a particular response then the actual figure might vary between 27.2% and 32.8% for the whole gathering. If a score was 70%, then you can say with certainty that the actual result (i.e. representative of all the delegates) is between 63-77%.
The above margin of error, even if applied to the full in every case, would not have materially changed the overall conclusions and reports of the survey as published in the Belfast Telegraph.
LucidTalk reviewed all the questions in advance to ensure they met professional polling standards in terms of clarity. The questions were non-leading and met the criteria required in a professional opinion poll.
LucidTalk wase represented at the conference as observers. We are satisfied that the 50 people who took part in the survey were wearing identification cards, issued by Sinn Fein on registration. These same identification cards were also used by delegates to take part in voting within the conference itself.
Bill White is managing director of LucidTalk. The Belfast Telegraph carried out its ard fheis survey in association with LucidTalk