Guys,
I just finished a stat class as a part of my training, and to solidity my knowledge, I played with the real data - my GMAT scores from the prep tests.. The discovery is somewhat shocking - the mean of my tests (686.9) is the same as my actual test (690). On a positive note, the score 780 is within a normal (Bell Curve) distribution for me, that is why I am going to retake the test in a couple of months.
I will continue my G-stat study, plotting separatedly Verbal and Quant scores and looking for any relationships and interesting conclusions.
DATA:
plotted chronologically, majority of tests are GMATprep,
MGMAT and a couple of Kaplans. However, I removed from calculations 1 Kaplan test where I scored 800 because that is non-sense.
METHODOLOGY: Guys, that is not the NASA, therefore data and findings will be sometimes awkward and even controversary:) Yet I want to make some conclusions or prove theories while having fun and solidifying my stat knowledge. I use Minitab 15 stat software to test my own GMAT data via different techniques: regression analysis, control chart etc.
What you see below is the CONTROL CHART, built by putting chronologically my GMAT scores from the very first (when I was totally blank) to the very last test (what is my real test 2 weeks ago). The UCL and LCL stand for Upper and Lower control limits, which are MEAN +/- 3 Q (sigma or standard deviation). As per stat rules, the 98% of the process data shall fit within these limits IF a process is statistically correct. However, if you'll see some weird points above or below the limits, then smth was wrong. The upper graph is test results themselves, the bottom graph is a moving range - difference between the two adjacent points. More about control charts
https://en.wikipedia.org/wiki/Natural_process_variationWhat the graph below shows you is that my educational process is
stable and predictable: with 98% confidence my results will be within the limits
HOWEVER... bear in mind my brain is not a Toyota factory; there are various factors (weather, mood, sugar level in blood) involved, each of which may shift my performance either way. Not to mention that you actually mix apples, organges, and smth else by putting together data from GMATprep,
MGMAT, and Kaplan tests. Yet, even such graph is better than nothing.
Once I get home, I'll make a regression graph between by Quant progress and scores in
GMATClub tests
Attachments
gmat.PNG [ 11.24 KiB | Viewed 1440 times ]