Have done some calculations on the bands. Here Goes. Assuming that the level 2 population forms the normal distribution in its performance on the test, and given the facts that the pass %age was 46%, band 10 forms the population ranging from p values of 0.486 and 0.540, band 9 forms p values from 0.432 to 0.486 and so on. Assuming a MPS of 66.67%, std dev of 14 (i.e 66% of the people scored between 52% and 82% ;fair?); the following are the conclusions. Passing score of 66.67%=>80 questions correct out of 120 Band 10 ; between 78 and 79 correct. Band 9; Between 76 and 77 correct. Band 8; Between 73, 74 and 75 correct. That means so many have missed by a one or two questions only. Your thoughts. … I dont mean to fret over how close anybody was but my question is with so little of the syllabus being tested shouldnt the exam be more representative of the size of the syllabus?

I doubt the score was 66%, from one of CFAI’s web postings it was in the high 50’s one year, believe it or not.

I’m in 6

6 also is not as far it seems… keep at it.

I calculated the worst score possible that I could get based on the number of questions in each area. For instance, an area assigned 18 points will have 6 questions. If my asterisk was between 50 and 70, I would score a 66.67% (4 correct. If you got 3 or less questions correct, you would be in the <=50 column). If you got 5 or 6 correct, you would be in the >70 column and assign yourself an 83.3% (5 out of 6). I assigned a 0% if less than 50. Does that make sense? Anyway, my WORST possible score is slightly over 60% and I did not pass and was in band 9.

This is really interesting jaspreet.

great analysis, although it makes me sick to think all I needed was 4 more points. FCK.

Tmeister B Wrote: ------------------------------------------------------- > I calculated the worst score possible that I could > get based on the number of questions in each area. > For instance, an area assigned 18 points will > have 6 questions. If my asterisk was between 50 > and 70, I would score a 66.67% (4 correct. If you > got 3 or less questions correct, you would be in > the <=50 column). If you got 5 or 6 correct, you > would be in the >70 column and assign yourself an > 83.3% (5 out of 6). I assigned a 0% if less than > 50. Does that make sense? Anyway, my WORST > possible score is slightly over 60% and I did not > pass and was in band 9. This absolutely make sense. btw what was your score ?

jaspreet, thanks for running those numbers and posting them. I was on the fence on whether or not I should submit a retabulation to CFAI. I failed and was in band 10. In fact, someone with my exact same break down on each section passed so i knew I was close and if your numbers are even somewhat accurate, I may have failed by a question or two. I figure, $100 bucks (for the retab) for the possibility of there being an error that results in me saving a whole year in studying is well worth it; worst case scenario i lose $100. Thanks again, and if I happen to pass, you can bill me for a round of shots! good luck on your studies.

Not to be pessimistic, I read from other posts that it is pretty worthless to get a retab in level 1 and 2.

Your right Nike, retab is pretty much worthless for level 1 and level 2 however as someone who scored in the 10 band and failed. I would not be able to study properly until I know beyond a doubt all possible routes had been taken if I’m going to slave away another year. So I think most do the retab for closure and in that way it is very useful.

Whoaa… I always like it when people use the statistics they learned, but that wasn’t a very good analysis. First, normality is the distribution of entropy, randomness, averaging and diffusion. It shows up everywhere in finance (maybe inappropriately) because there are elegant theorems about arbitrage-free markets, Markov processes, etc that imply normality. There is a very useful stochastic calculus based on it that makes lots of pricing possible and maybe easy. That doesn’t mean that the distribution of a bunch of finance test scores is normal (it’s almost certainly better approximated by a beta-binomial). If you assume normality, you are implying lots more randomness in the process than exists. What really happens is that there are people with different levels of preparation, intelligence, # previous times taking the test, etc… and each person begins the test with some different score distribution. Their score on the test is the number of questions they get right and it’s not too much of a stretch to say the number they get right is binomial depending on an individual parameter p. For a bunch of technical reasons, p is usually assumed to have a beta distribution. That means scores are much more like beta-binomial than normal. The big difference is that normality packs lots of observations near the mean. Beta-binomials (and most other standard distributions) don’t. One cheesy hand-wavy way of explaining that is to compare the Chebyshev limits with the normal distribution - big difference. Based on the distribution above, what’s the probability that someone got a 15? (ans: probably 1 person in the world would have score this poorly) In real-life, how many people in your test center got a 15? I’ve graded plenty of LII exams that got 15% or less (they are easy to grade incidentally). The bands are almost certainly much wider than suggested by the analysis above. In any event, I am still a big seller on retab, even for LIII. If $100 doesn’t mean anything to you, send it to me.

BTW - “I dont mean to fret over how close anybody was but my question is with so little of the syllabus being tested shouldnt the exam be more representative of the size of the syllabus?” On another lesson topic from stat class - the validity and reliability of your score doesn’t depend on the size of the universe you select questions from, it depends on the number of questions on the test. That’s why a random sample of 100 people do approximately as good a job in predicting the winner of the Presidential contest as they do in predicting the next mayor of Stamford.

jaspreet Wrote: ------------------------------------------------------- > Have done some calculations on the bands. > > Here Goes. > > Assuming that the level 2 population forms the > normal distribution in its performance on the > test, and given the facts that the pass %age was > 46%, band 10 forms the population ranging from p > values of 0.486 and 0.540, band 9 forms p values > from 0.432 to 0.486 and so on. > > Assuming a MPS of 66.67%, std dev of 14 (i.e 66% > of the people scored between 52% and 82% ;fair?); > the following are the conclusions. > > Passing score of 66.67%=>80 questions correct out > of 120 > > Band 10 ; between 78 and 79 correct. > Band 9; Between 76 and 77 correct. > Band 8; Between 73, 74 and 75 correct. > > That means so many have missed by a one or two > questions only. > > Your thoughts. … > > I dont mean to fret over how close anybody was but > my question is with so little of the syllabus > being tested shouldnt the exam be more > representative of the size of the syllabus? jaspreet, thanks for the analysis…makes me more pissed off that i could have passed by a couple of points!!! =) regards, “the band 10 loser”

I guess people believe what they want to believe.

JoeyDVivre Wrote: ------------------------------------------------------- > I guess people believe what they want to believe. haha. Gota love noytkimcap totally ignoring you’re own analysis.

jalmy8 Wrote: ------------------------------------------------------- > JoeyDVivre Wrote: > -------------------------------------------------- > ----- > > I guess people believe what they want to > believe. > > > haha. Gota love noytkimcap totally ignoring > you’re own analysis. yes its called “selective reasoning to make you feel better”…hope it’ll work for you…