Thursday, 19 July 2012

Further EQE paper C 2012 statistics

If Pete can make nice figures, so can I.
Some additional statistics for paper C:
  • 1410 candidates scored above 0
  • the average score was 41.56
  • 467 candidates (33.12%) passed the exam
  • 157 candidates (11.13%) scored a compensable fail
  • 786 candidates (55.74%) failed

    For those who passed: Congratulations!
    For those who failed: We have some excellent courses that I would like to recommend.
    On an unrelated note: Rome is a very nice city to visit.

  • 11 comments:

    1. So the stats in the post below (only 33% pass) are wrong? Damn.

      ReplyDelete
      Replies
      1. Sorry. Completely wrong copy-pasting. I will correct this immediately.

        Delete
    2. One should explain me why they can always (i.e. each year) have a kind of Gaussian curve around 40-50 marks

      ReplyDelete
      Replies
      1. Because with large numbers of candidates, most statistical processes tend to a Gaussian (see the Central Limit Theorem), while the papers are similar as between years, as is the mode of preparation of the students?

        Delete
      2. Then, why the significant dip between 40 and 50 marks? It is always there, and is alien to all Gaussian curves.

        Delete
      3. Some of the papers are discussed more than other papers. Of course, there is no use in long discussions about a paper that's worth 20 or 70 marks.

        For papers marked in the ranges 43-46 and 48-51 it is very important that the 'correct' amount of marks is awarded. After the discussion some rating will not be changed, and some will be raised or lowered. As a result, you see a small dip in the compensable fail range. When you zoom in on the stats, you will see that only a few candidates get 44 or 49 marks.

        Delete
      4. I understand this is a very valid concern.

        After all, no one wants to fail unnecessarily.

        Still, a review that kicks out grades to a different bin, either above or below, can only show that there is some fuzziness in the evaluation, at least 2-3% in either direction.

        Delete
    3. There will always be a tendency towards a double Gaussian curve. The 'lower' curve is for candidates that have it wrong; the higher for candidates that have it right. In the past we have seen this very clearly for A exams with a lot of marks for the main independent claim. If the exam has many independent questions, the picture will blur. Nevertheless, candidates who are well prepared and fast will score many high Gaussians and in total will contribute to the higher part of the curve. Similarly, candidates who are not well prepared and slow will contribute to the lower part of the curve. So, our analysis of the past decades shows those two peaks and a dip in the middle. Sometimes, the questions in the exam have a (too?) high correlation (if you do the first bit wrong, you automatically do more wrong; or the other way around) and the peaks get very clear, just as the dip.

      ReplyDelete
      Replies
      1. I am still curious to understand why/how the curve is always center on 40-50 and not on 30-50 (bad year) or 50-60(good year). As well as why/how can the passing rate be always aroun 30-35%. Are the candidates and their preparation so homgenous from one year to the other?

        Delete
    4. The passing rate actually fluctuates rather a lot. For paper C the last decade it was between 32% and 53% for the first time sitters (similar for D). The exam committee does its best to keep the passing rate stable, but does not fully succeed. In a first marking round the committee tries to identify 'unexpected' answers and sometimes adapt the marking accordingly. Yes, I think that the group of candidates is big enough to give stability. These issues explain why the center usually is somewhere between 45 and 50.

      ReplyDelete
      Replies
      1. Thank you for answering

        Delete