> pierre dorgueil wrote: > > Wilma Silbermann a écrit : > > > > > > Finally, with regard to our exam scores - please see our faq; > > > http://www.lpi.org/faq.html#Exams13 > > > > Sorry, but I did go to this FAQ, and don't understand anything more now. Note > > that the arrays are empty on this page... Yes, sorry, I did not have the numbers at the time when I wrote the FAQ. Scott Murray left before he worked out those derived statistics, and his successors, Kara Pritchard and Alan Mead, have other priorities now. > > More precisely, when it is stated that the exams "range from 200 to 800", does > > it mean that: > > > > - the minimum available note (all false) is 200, and the maximum (all correct) > > 800, each question having a variable weight. So the actual required passing > > score of 460 would be about 43% of the range, whatever the note means? > > - the maximum note is variable between 200 and 800, so no info at all is > > available about my own performance (should get 460, got 653, but nothing more). > > The exam seem to be scaled with a mean beta note of 500 (50%?), so could I > > consider that my 460 is 50% too, leading to a maximum note for my exam of 736? > > - anything even more complicated... Just what it says in the FAQ: People who took the exam in the beta period had some range of scores. Scott computed the mean and the standard deviation of all these scores. Then he scaled raw scores such that the mean became 500, and the standard deviation 100. Assuming that the distribution is normal, then the range of 200 to 800 means: [mean +/- 3 standard deviations], and that covers 99.5% of the normal distribution curve. Effectively, no-one gets a score outside that range. Then some cut score was defined, based on the judgement of a team of experts on the difficulty of the test items, and on the actual performance of test takers on the test. The cut score for 101 was set to ..err.. 490 I believe: this implies that only slightly more than half of the people who took the beta test, passed. If you score over 500, you are better than average (=50% of beta testers). A score of 653 means you are better than 94% of the beta testers. Note that this is not a linear scale, it follows the normal distribution. > > This should of course be of no special interest for me, provided I passed. But I > > would like to insist on understanding this for the following reason: > > > > I was certified RHCE/RHCX 1.5 year ago. I then gave official RH certifications > > in France. Now that I leaved RedHat, although I continue giving official RH > > courses (but no more certifications), I would like to give also MandrakeSoft and > > SuSE training courses, that are both LPIC-targeted. I would like to be able to > > understand as much internals as possible (I'm of course aware of any > > confidentiality reasons), simply to be able to do correct pre-exam or > > post-failed exam support for my future students that will also be future LPIC > > attendees. > > > > I can already evaluate the LPI-102 as slightly more difficult than the MCQ-part > > (that counts for 1/6 of the time and 1/3 of the note) of RHCE exam. The > > questions seem correctly written and reasonably unambiguous in both cases. But > > the mean score in RHCE case is 80% (unscaled), and at least 50% is mandatory. > > What about LPIC? Again, apologies if this info is present on your site, but I > > didn't find it. The important point is, that the question of what percentage of questions you answered correctly is improper. % correct is a bad measure, because it also depends on how good or difficult the questions are, and we do not know that beforehand. Our way of computing the score compares you directly with your peers, and this relative assessment is much more informative than the absolute number. The latter is actually relative to the difficulty of the test questions, which is not well known before one tries the exams with real candidates. > > Furthermore, in LPIC case, is there any explanation of a good way to work after > > a failed exam, or do you consider that the relative opacity (in my opinion, > > perharps your site is simply too difficult to approach for non-native english > > speakers?) that surrounds this exam must be taken as a guarantee of validity? This relates to 2 different things: 1) after the exam, you do not get information on what questions you answered wrong. It is bad for the confidentiality of the test to repeat in detail what is wrong. However, it is achievable to give an assessment per subject area that may tell you what to concentrate on for further study. We are discussing with VUE how to implement this, and it may be introduced some time in the future when the exams get renewed. 2) LPI does not care how you obtain your expertise, in fact we try to remain neutral with respect to training institutions, so we have little information on training on our site. However, this page contains all relevant links: http://www.lpi.org/c-preparation.html > > I think honestly these two exams are comparable as they both aim to guarantee > > linux abilities. They only disagree on the best way to do this, and neither is > > totally wrong in my opinion. I assume that with the "two exams" you refer to the LPI and Red Hat certification exams. The reason why we use multiple choice tests exclusively instead of a "hands-on" test like Red Hat, is explained on: http://www.lpi.org/MCvsHandsOn.html -- Tom Peters Director of the Board & Exam Development Specialist, Linux Professional Institute e-mail: [EMAIL PROTECTED] -- This message was sent from the lpi-examdev mailing list. Send `unsubscribe lpi-examdev' in the subject to [EMAIL PROTECTED] to leave the list.
