Hi Ian,
I am extremely sorry
for the typo in option B… Actually I ended up typing the entire problem. Hence, the error in typing the extra zeroes. There was a typo in C too. It is 74% and not 84% .With due regards, I have corrected the errors in the original post.
Overall, I too found the problem kinda’ ……sensible but pretty challenging as the language in the stimulus and the options is very convoluted, which makes it extremely hard to interpret.
Here’s my analysis:
Sentence 1 (Background Info): CGI conducts MAT, introduced ten years back (an online-based aptitude test).
Sentence 2 (1st premise): Latest statistics released by CGI show that the average error rate, a statistical figure to assess the difficulty level of a question and determined only after the number of respondents reaches at least thousand, of ten toughest questions in a sample group of MAT tests conducted within any pre-defined time interval not earlier than three years ago has never exceeded 70%.
My inference: Error rate is a statistical figure to judge the level of difficulty. This figure, however cannot be constant for any question and will keep changing as more and more students attempt it. The error rate may increase or may decrease depending upon whether more students answer it correctly or not. This means that a question that was in the top ten tough question list in a particular time – frame MAY OR MAY NOT retain its position. Point is that owing to the dynamic nature of the error rate, we have a different combination of ten toughest questions for different time-frames of 1 month, 6 month, 1 year, 2 years or even 3 years (i.e. for different time intervals). Now, statistics indicate that in none of these different sets of toughest questions has the AVERAGE error rate climbed over 70%. Although, the individual error rate of the question at the top of the list could still be way above 70%, those of the others towards the bottom are so much lower than 70% to nullify the effect and bring down the average to less than 70%. Or in other words, the percentage of the total number of people answering the ten toughest questions wrong out of the total number of respondents attempting these has never exceeded 70%.
Sentence 3 (2nd premise): A particular question recently introduced by CGI in MAT was, however, found to have an error rate of more than 80%, attempted by approximately 5000 candidates worldwide.
My inference: A particular question introduced was attempted by 5000 respondents, out of which more than 4000 got it wrong. Note that the sample population of 5000 was well-represented as indicated by ‘worldwide’, i.e. the population belonged to different demographics
Sentence 4 (Judgment/Opinion/Intermediate conclusion): Hence, this question could be considered among the toughest of the questions to have appeared on the MAT
My inference: The fact that less than 1000 out of 5000 students could only manage to answer the question right INDICATES (not PROVES) that it is among the toughest. (i.e. it could potentially count amongst the top ten toughest questions for a statistical analysis yet to be carried out as it is just recently introduced.)
Sentence 5 (Counter premise): But, some respondents contend that the new question was indeed so tough need not be true as it is also quite likely that the official answer to the question itself is wrong.
My inference:
The word ‘but’ initiates a debate with some respondents contending that the answer itself perhaps could be wrong and is a point of contention, referenced by the question stem. The sentence introduces another potential reason why the new question may probably have such a high error rate.
Question Stem: Which of the following, if true, would most help to arbitrate the contention presented in the argument?We need to find an option that produces additional evidence to either support the contention or destroy it.
A.
Wrong. Even if 4750 respondents who answered the question right scored 99 percentile on MAT overall, this does not really resolve the contention whether the answer is right or not. It could be quite possible that these respondents have actually marked an answer, which is actually wrong and all the 250 remaining respondents who got it wrong have marked an answer which is actually correct and also have scored 99 percentile or perhaps even higher than the 4750 respondents who got it right.
B.
Wrong…Looks to be a red herring, anyways does not solve the contention of the debatable answer. What I could really make out is that this is a percentage trap. In essence, the option says that the maximum error rate of the ten toughest questions within any time band recorded until date is 82.5% and the minimum error rate for these is 60%.
Quite likely that the question which achieved an error rate of 82.5% actually had the correct answer marked. Hence, this in no way proves that a question with an error rate of 80% or above is necessarily wrong.
C.
Wrong…Another distracter
(I later began thinking that this question is testing quant more than verbal or anything else related to CR) 
…Going ahead, this option also does not resolve the contention whether the answer is right or not. In fact, the fact that the avg. error rate was 74% need not mean that none of them were above 80%. Quite likely that one of the ten questions had a 90% error rate while the others had way below 80% to bring the avg. to 74%. Proceeding ahead, the 90% question can be dismissed as right with the same reasoning as in B, thus intensifying the contention further.
D.
Wrong. Same as B and states the reasoning explicitly to eliminate D.
E.
Correct. It states that studies by certified psychometricians
(those dealing with measurement of human intellect vide IQ and aptitude tests, or those who dealing with measurement of human traits and attitudes, etc….etc) state that any question can never achieve a max. error rate not lower than 80% (i.e. 80% or more), even after allowing for a max. grace variation of up to 5%. This means that no matter how difficult a question is, if it is presented to a well-represented respondent sample population of at least 1000, you will find that generally the toughest question will be answered wrong by at most 75% i.e. 750 people, not more. However, since the error rate is subject to fluctuations (as discussed previously), it will never exceed 79.99999……999%. (i.e. in any case, at least 201 respondents out of 1000 have to answer the question correctly, or else there is something wrong with the question). This proves that the answer to the new question has to be necessarily wrong as the error rate is more than 80% and resolves the contention
close
Updated on: 30 Sep 2014, 05:15
69%
(03:04)
wrong 
