PDA

View Full Version : Your Score

Sew Lining
12-27-2005, 12:46 AM
Finally got my score. If you failed, as I did, please calculate an estimate of your total points and post your results in the following format:
SCORE / MIN POINTS / EST POINTS / MAX POINTS.

Calculation:
For the minimum (maximum) possible score, use the bottom (top) of each range. My minimum possible score is 59.0. Similarly my max possible score is 76.4. The estimated points would be the average of the two.
(edited b/c my brain is so fried from the exams that I'm a moran and make easy things difficult)

So I would report:
5 / 59.0 / 67.7 / 76.4

This tells us that the minimum passing score was definitely higher than 59.

tommie frazier
12-27-2005, 01:12 PM
according to the analysis, i figure I got between a 56 and a 68.

scored between 95% and 115% of average.

my guess is I missed by about 3 or 4 points, tops. rats.

5/56/xx/68

I also restricted my point totals per question to multiples of 0.25.

sew-you look closer than I was.

Sew Lining
12-27-2005, 01:53 PM
sew-you look closer than I was.

Horseshoes or hand grenades, anyone? :D

If other folks care to do the calculations and post the results, I promise to make some real pretty graphs and calculate some neato statistics!

Bullfrog1220
12-27-2005, 02:07 PM
4/52/xx/70.5

In my very quick "post-exam analysis", I figured that I got between 59 and 62 and ended up with a 4. I'm looking forward to the colorful graphs.

The Continental
12-27-2005, 02:17 PM
5 / 63.5 / xx / 82.25

That should help move the min pass mark a bit. @#&#%!

Sew Lining
12-27-2005, 02:33 PM
5 / 63.5 / xx / 82.25

That should help move the min pass mark a bit. @#&#%!
Holy cow!

That 63.5 came from taking the minimum on each and every question? That's amazing!

Keep your chin up, we'll get it next time.

sd978
12-27-2005, 03:48 PM
Trend certainly looks weird on this one. I better get ready to score at least 70 in this one next time. Just being on the ballpark or trying to get closer to the 60s is not enough given the way this goes.

The Continental
12-27-2005, 05:13 PM
Holy cow!

That 63.5 came from taking the minimum on each and every question? That's amazing!

Keep your chin up, we'll get it next time.

Yeah, at least it confirms I was close. That much less to do in a year...

Sew Lining
12-27-2005, 07:07 PM
Okay, from the results posted so far, it looks like the pass mark is:

MIN = 63.5 = Continental Min
EST = 68.5 = Avg [Tommie Max, Min, Bullfrog Max, Min]/0.9
MAX = 77.0 = Avg [Tommie Max, Bullfrog Max]/0.9

Tommie, can you go back and calc your max and min based strictly on the range boundaries?

Have I missed anything? Can anyone refine this from what we have so far?

GefilteFish144
12-27-2005, 09:22 PM
Grade: 5 (59.5/xx/71.5). Point estimate is 62.5, but that's probably a bit conservative.

Interesting sidenote: Question 21 -- range of actual 25-49%; raw/mean >175%. So what they are basically saying is that they are still giving a ridiculous passmark despite the fact that most candidates bombed a 4.5 point question. :shake:

tommie frazier
12-27-2005, 11:06 PM
If I redo it not restricted to the multiple of .25 for the score, i am:

5/55.38/64.92/74.46

if restricted as above, I get:
5/56/62.13/68.25

Sew Lining
12-28-2005, 01:27 AM
Thanks, Tommie.

Okee dokee, here we are with some more "neato statistics":

MIN = 63.5 = Continental Min
EST = 70.09 = Avg [Tommie Avg, Bullfrog Avg]/0.9
EST = 71.31 = Avg[Sew Avg, Tommie Avg, Continental Avg, GFish Avg]/0.95
MAX = 77.0 = Avg [Tommie Max, Bullfrog Max]/0.9

BIAS:
1. Slight downward due to using the top of each range as 24, 49, 74, 99% instead of 24.99, 49.99, 74.99, 99.99%. Max amount of bias = 1%. Might need to add 1 point to the estimates.
2. Significant upward due to assuming the mean of the range 0-24% is 12%, when it will be a bit lower due to some questions being left blank. Max amount of bias due to this, for folks with a 5, should be about 0.04.

(Assume the pass mark is 70 as shown above. Then the lowest possible 5 score corresponds to 70*0.9=63 points. Assume this person got 100% on 63 points and left 37 points blank. The 37 blank points would contribute 0.12*0.37=0.0444 to his estimated score when using my estimation method, but of course that'd be wrong b/c he actually left those points blank)

Soooooo, to revise the above, we have:

EST = 70.09 +1/-4 => 66.09 - 71.09
EST = 71.31 +1/-4=> 67.31 - 72.31

Let's just say it's between 66 and 72. Pretty darn high. As always, more data points help with the estimation - doesn't matter what you scored, as long as you have an analysis to look at. Low 5's and high 4's are especially welcome.

Someone please correct me if I've goofed up somewhere.

tommie frazier
12-28-2005, 12:22 PM
I still think that the most error comes from not limiting actual scores to quarter point values. No one gets 2.97 points. you might get 2.75, or 3, but not 2.97. in my post above, you can see that limiting to .25 multimples cuts over 6 points from the max score I could get.

Sew Lining
12-28-2005, 12:55 PM
I still think that the most error comes from not limiting actual scores to quarter point values. No one gets 2.97 points. you might get 2.75, or 3, but not 2.97. in my post above, you can see that limiting to .25 multimples cuts over 6 points from the max score I could get.

Is partial credit limited to quarter point values? Perhaps I should know this by now, but I've never really cared enough to find out.

tommie frazier
12-28-2005, 12:58 PM
Having graded papers before (not these exmas, but some papers), there comes a threshold at which you can't parse the things any finer. I picked quarter points here. I just can't imagine that they track tenths.

one more data point from a friend:
5/60.25/66.25/72.25 (uses the quarter point restriction)

Sew Lining
12-28-2005, 01:54 PM
I agree that quarter points seems like the most reasonable approach. I'm afraid that most people don't want to take the time to go question by question and assign quarter points.

Since most of the respondents have used the max/min method, I think it'd be best to have everyone report their points that way. I do agree that your method would be better, but I think folks generally won't go to the trouble of assigning quarter points.

When averging across all who report their scores, the bias involved in not assigning quarter points should be something like:

Question.....Est Avg..........No. of
...Point......Bias (Exam.....Questions
..Value........Points)
..1.5...........+0.055..............1
..2.0...........+0.115............12
..2.5...........+0.050..............2
..3.0...........+0.110............14
..3.5...........+0.045..............2
..4.0...........+0.105..............4
..4.5...........+0.040..............1

Tot Bias = Sumproduct[Bias, No. Questions] = 3.63 points

Note that this incorporates an adjustment for both of the bias sources I list in my previous post, so I'll just use +3.6 as the bias estimate.

Sound okay?

Sew Lining
12-28-2005, 02:38 PM
Here's a new est including Tommie's friend:

EST = 67.97 = 0.8*(Avg[Sew Avg, Tommie Avg, Continental Avg, GFish Avg] - 3.6)/0.95 + 0.2*Friend Avg / 0.95

I'm thinking this looks more and more like a pass mark of between 67 and 69.

Sew Lining
12-29-2005, 01:38 PM
Anyone else care to post their min/max score?

Come on, folks, don't make me obsess about my failure and the pass mark all by myself! ;)

GefilteFish144
12-29-2005, 02:33 PM
Shouldn't take that many samples, I don't think. When we did this exercise for Exam 8, it took just 3 of us to correctly project a pass mark of 46. Considering that my minimum was 44.5, you couldn't begin to imagine my frustration of having another 4 months of studying to improve my performance by 1-2 points. :swear:

Avi
12-29-2005, 02:36 PM
Anyone else care to post their min/max score?

Come on, folks, don't make me obsess about my failure and the pass mark all by myself! ;)Some of us do not have our report yet :(

Wilson2
12-29-2005, 03:15 PM
5/61.25/xx/79.04

Super Silver Haze
12-29-2005, 04:32 PM
5/62.5/xx/78.34

Sew Lining
12-29-2005, 06:09 PM
Update:

EST = 68.6 = (1/0.95)* [ (1/7)Friend + (6/7){Avg[Sew, Tommie, Continental, GFish, Silver, Wilson]-3.6} ]

I'm assuming the folks who scored a 5 are symmetrically distributed about the point MIN PASS * 0.95.

This seems the most reasonable assumption to me, but could see how it's possible to have more folks bunched up at the high 5's versus the low 5's.

Also, an argument could be made that the bias adjustment of 3.6 should be a little higher. If it were 4, the estimate would be 68.24.

Pack Fan
01-03-2006, 09:00 AM
5/57.1/xx/76.1

Sew Lining
01-03-2006, 11:33 AM
Including Pack:

EST = 68.32 (assuming bias = 3.6)
EST = 67.95 (assuming bias = 4.0)

Avi
01-05-2006, 12:36 PM
OK I have 3 sets of 3 estimates. Low uses the bottom of the range (0, .25, .5, .75, or 1), High uses the top of the range (.24, .49, .74, .99, or 1), and Mid uses the average.

The first set of 3 is the straight sumproduct of the point value with the number of points. The second set of 3 is has each question rounded to the nearest .25. The last set of 3 has each question rounded DOWN to the nearest .25.

Score: 4

Standard: 47.625/58.065/68.505
Rounded: 47.75/57/69
RoundFloor: 47.5/52.75/62

MathGuy
01-06-2006, 11:30 AM
Grade: 5 (59.5/xx/71.5). Point estimate is 62.5, but that's probably a bit conservative.

Interesting sidenote: Question 21 -- range of actual 25-49%; raw/mean >175%. So what they are basically saying is that they are still giving a ridiculous passmark despite the fact that most candidates bombed a 4.5 point question. :shake:
So you think that because most people bombed a 4.5 point question, that more people should pass?

Tri4Ben
01-09-2006, 05:11 PM
5/54.63/74.67

Now it looks like I wasn't really all that close compared to some of you guys.