Scioly.org

pikachu4919 wrote: Wait kenniky why were you at home during nationals this year...?

check the 2017 MA thread

Did anyone else take to Div C Rocks & Minerals test?

bernard wrote:Here's some data you can play with. I've included the Pearson correlation coefficient comparing each event to overall team rank. This can be done for any tournament's results.

So I'm not really understanding this. Do the numbers get higher if more good teams succeed in a certain event? (If so, WIDI is almost like the lottery )

Tailsfan101 wrote: So I'm not really understanding this. Do the numbers get higher if more good teams succeed in a certain event? (If so, WIDI is almost like the lottery )

Coefficient of correlation tests the correlation between two variables, in this case overall team ranking and team ranking in a certain event. The closer the coefficient of correlation gets to +1.00, the closer to a perfect correlation between overall team rank and the team rank in a certain event.
You're right about WIDI, and it's likely because it's difficult to perform consistently, so alot of teams just gave up. Ours definitely did.

Uber wrote:

Tailsfan101 wrote: So I'm not really understanding this. Do the numbers get higher if more good teams succeed in a certain event? (If so, WIDI is almost like the lottery :lol:)

Coefficient of correlation tests the correlation between two variables, in this case overall team ranking and team ranking in a certain event. The closer the coefficient of correlation gets to +1.00, the closer to a perfect correlation between overall team rank and the team rank in a certain event.
You're right about WIDI, and it's likely because it's difficult to perform consistently, so alot of teams just gave up. Ours definitely did.

Pearson correlation coefficient ranges from -1 to +1, where -1 indicates negative correlation (teams that do well overall do poorly in this event), 0 indicates no correlation, and +1 indicates positive correlation (teams that do well overall do well in this event). Note PCC does not tell us the slope of a regression.

When the team rankings are inverted, the correlations for each event become negative. However, when you're in the scoring room the effect of an event scored backwards with the best teams ranked last may be more subtle because the overall team rank is calculated with the error so it's not as simple as an opposite sign.

Keep in mind, absolute correlation between individual events and the overall team ranking isn't really desirable. That would say there is no way for individuals on a team to outperform the overall team. You have to expect some variation. Low correlation may just say the talent for that event is more widely spread, or less dependent on team support. Not that the test isn't good.

In general correlation ONLY tells you things move together, NOT why.

Jeff Anderson
Livonia, MI

Hi jander14indoor,

However, I would argue for study events, a lower value might mean an easier test in general. For example, look at Anatomy and Physiology for both divisions. In Division B, the test was mostly on the easier side, allowing people from teams that don't do so well to have a chance to perform well in this event. However, in Division C, the test was mostly hard and well-written, which means the good teams would do well (just look at the top 6 in C; there are all from the top 10 overall, except for (me ) Ladue, who came in at 11th).

Based on this, I would expect the 2015-2016 Division B A&P value to be on the lower side as well, if anyone would care to prove my hypothesis right/wrong?

Thanks,
sciolyFTW_aku

From what I have gathered, the common thread for Anatomy is that ES Patty Palmietto writes tests that are generally too easy and fail to separate out strong teams from weak teams. I know she was the Div B ES at nationals this year and I have taken tests from her in the past. They were generally focused on very basic anatomy or process skills, which therefore means in-depth knowledge of Anatomy is of limited benefit. The only area she tends to go into detail on is disease identification, often based off of images.

chalker wrote:

chalker wrote:
Max event correlation: 0.52
Average event correlation: 0.39
Standard deviation: 0.09
Min event correlation: 0.20

Ecology had a correlation coefficient of 0.28. This means it was about 1 standard deviation below the average correlation for events. There were several events with lower correlations.

In essence, what this means is that statistically, the resulting ranks in Ecology are reasonably well aligned across all teams with the overall team ranks.

Oops.. I did this during lunch and thanks to Bernard posting his sheet I realized I did the wrong ranges. Below are the actual numbers, but the general conclusion is the same:

Max: 0.85
Average: 0.75
Std. Dev.: 0.10
Min: 0.42
Ecology: 0.57

WIDI and EV are lower than Ecology.
8 of the 23 events are under the average.

Chalker, is there ever any discussion of removing or refining events if they consistently show a poor correlation with team scores? If WIDI is so variable and a poor predictor of team outcome, why has its grading/scope not been better standardized or controlled?

Uber wrote:
5. MIT ecology was hands down the most brutal test I've ever taken. Golden Gate also had a much more difficult test than nationals, with more critical thinking involved. We won both. The national test came nowhere close. We finished most stations and double-checked with time to spare.

.

I'm glad you liked my MIT test. It was supposed to be brutal and make sure to separate the good teams from the weak ones. I remember you won first by a significant margin, so congrats!