|All Reviews > sjas31's Reviews|
Joined: Dec 16, 2018
I got these 7 CAT's in $19 instead of $49 and I am quite happy with the money spent on them. I have used only 2 of them till now. They are great in training and building stamina. After the test, you get good analytics and bifurcation of questions with level, time spent and review option. It gives you a rough idea where you are and how you are tackling different level questions. Though I have some doubt in the algorithm on which it worked but that's not the purpose of purchasing them. It worked differently for consecutive incorrect and correct attempts. Anyway, let's not go to endless algorithm discussions, it's worth spending the money and recommend to others.
Glad to hear you're finding the Veritas Prep tests valuable! Hey, since you mentioned the algorithm and that it reacts differently even under the same correct/incorrect response patterns, I figured I'd chime in with an explanation. We use Item Response Theory - the same adaptive model that the official GMAT uses ...we actually had one of the former GMAC psychometricians consult for us on our system - on our tests. IRT adapts using three parameters - the "difficulty" metric (the b-parameter), the "reliability" metric (the a-parameter - basically a gauge for how strong each question is as a predictor of ability), and the "guessing" parameter (the c-parameter, which factors in how likely it is that a correct answer was a guess vs. actual knowledge/ability). The system is designed to serve the next question by looking for the question that's most likely to give it the most possible information about your ability, and it gauges that using the A/B/C parameters. *AND* it's also bound by content balancing (so that you see a representative number of DS vs. PS questions, for example, and word problems vs. geometry problems, etc.). Plus we use a handful of unscored, experimental questions just like the official test, too. So the system doesn't always behave the way you'd expect it to based one some of the "get it right --> harder question; get it wrong --> easier" YouTube/blog tutorials on CATs would want it to, but we think that's a feature and not a bug...we're serving it up using the same IRT framework that the official test uses.
Thank you so much for helping me in understanding the algorithm tech. I appreciate your post. I did not received notification of your comment, sorry for bring late in replying. Have a good day!