0 members (),
130
guests, and
40
robots. |
Key:
Admin,
Global Mod,
Mod
|
|
S |
M |
T |
W |
T |
F |
S |
|
|
|
|
|
|
1
|
2
|
3
|
4
|
5
|
6
|
7
|
8
|
9
|
10
|
11
|
12
|
13
|
14
|
15
|
16
|
17
|
18
|
19
|
20
|
21
|
22
|
23
|
24
|
25
|
26
|
27
|
28
|
29
|
30
|
31
|
|
|
|
|
|
|
|
Joined: May 2013
Posts: 2,157
Member
|
OP
Member
Joined: May 2013
Posts: 2,157 |
Wondering if anyone can explain (to a non-mathy person) what the difference is between a national percentile ranking and a normal curve equivalent. I'm looking at score reports for my kids and for DS for math and reading it is 99 across the board for both math and reading NPRs and NCEs, but for DD, the NCE scores are quite a bit lower than the NPRs in both math and reading.
|
|
|
|
Joined: Apr 2014
Posts: 4,078 Likes: 8
Member
|
Member
Joined: Apr 2014
Posts: 4,078 Likes: 8 |
NCEs are another transformation of z scores (standard deviations), just done in such a way that (in a normally-distributed population) a NCE value of 1 = percentile 1, NCE 50 = 50th %ile, and NCE 99 = 99th %ile. Each NCE unit is a fraction of a standard deviation.
Percentiles are an ordinal measure, which describe your ranking in the norm group.
If you have a non-normal distribution of scores, the NCE and percentile will line up even less well.
...pronounced like the long vowel and first letter of the alphabet...
|
|
|
|
Joined: May 2013
Posts: 2,157
Member
|
OP
Member
Joined: May 2013
Posts: 2,157 |
Ok, so I'm kind of slow with this. I'm assuming that the percentile ranking is more meaningful than the NCE? And what does it mean if they are the same with DS (everything is a 99) but with DD the NCEs are a lot lower (5 to 10 points)? So for instance for math her percentile ranking was 98 (she was unmedicated and her scores dropped a lot), but NCE was more like 90 or 93.
|
|
|
|
Joined: Apr 2014
Posts: 4,078 Likes: 8
Member
|
Member
Joined: Apr 2014
Posts: 4,078 Likes: 8 |
NCEs are just a different measure. Percentiles are really bunched up around the middle of the curve, whereas NCEs are equal-interval measures (based, as they are, on the standard deviation), so they're more spread out than percentiles are in the middle, but closer together than they are at the extremes. That's why out at the tails, a percentile of 98 has a lower NCE. Everything is 99 for DS because NCEs of 99 are defined to be at the 99th %ile. Percentiles are easier for most non-mathy people to understand, as they represent your place in line. The NCE is exactly as meaningful as any other transformation of the standard deviation/z score, such as the standard score used for IQs and achievement tests, or scaled score used for IQ subtest scores. It tells you your distance from the center of the bell curve (defined as NCE = 50). A NCE is like a standard score where the mean = 50, and the standard deviation = about 21.
It's value is that, as an equal-interval scale, it is perceived as being better for making comparisons between different measures, and for measuring growth.
...pronounced like the long vowel and first letter of the alphabet...
|
|
|
|
Joined: Dec 2012
Posts: 2,035
Member
|
Member
Joined: Dec 2012
Posts: 2,035 |
Percentiles can be a bit misleading. If half the class of 31 got 100% then the 16th kid who got 99% would be at about the 50th percentile. The difference in the scores suggests there was something like that going on. When they somehow? Forced the results into a normal distribution some kids went down. They may have added an age rather than grade rating - eg 100% at 7.8 beats 100% at 7.9 etc.
Ok I cross posted. Disregard the second part of the post.
Last edited by puffin; 12/03/14 09:18 PM.
|
|
|
|
Joined: Apr 2014
Posts: 4,078 Likes: 8
Member
|
Member
Joined: Apr 2014
Posts: 4,078 Likes: 8 |
No, puffin, actually NCE 92 or 93 is always going to be 98th %ile, in a normally-distributed population. It isn't a peculiarity of this specific norm group. There will be some oddities comparing NCE to %ile if the data set is not normally-distributed, though.
Although I agree that %iles can be misleading, especially in a tightly-bunched group with outliers. Just like medians. It's just rank-ordering.
...pronounced like the long vowel and first letter of the alphabet...
|
|
|
|
Joined: May 2013
Posts: 2,157
Member
|
OP
Member
Joined: May 2013
Posts: 2,157 |
Dh majored in math, I'm going to have to show him this thread and make him draw me a picture.  Thanks for the explanations.
|
|
|
|
Joined: Feb 2012
Posts: 1,390
Member
|
Member
Joined: Feb 2012
Posts: 1,390 |
Here's the picture he'll draw you, blackcat (pdf file).
|
|
|
|
Joined: Dec 2012
Posts: 2,035
Member
|
Member
Joined: Dec 2012
Posts: 2,035 |
No, puffin, actually NCE 92 or 93 is always going to be 98th %ile, in a normally-distributed population. It isn't a peculiarity of this specific norm group. There will be some oddities comparing NCE to %ile if the data set is not normally-distributed, though.
Although I agree that %iles can be misleading, especially in a tightly-bunched group with outliers. Just like medians. It's just rank-ordering. I go that after I read the post that crossed mine. I just couldn't be bothered deleting the whole thing. I was thinking of something else. Next time i will check the book first. Sorry.
Last edited by puffin; 12/03/14 09:30 PM.
|
|
|
|
Joined: Apr 2014
Posts: 4,078 Likes: 8
Member
|
Member
Joined: Apr 2014
Posts: 4,078 Likes: 8 |
Here's another way of thinking about the NCE: If DD scored a NCE of 92, that's two standard deviations above the mean (50 + 21 + 21), which in standard score terms (like a WISC or WJ score) would be 130, which is at the 98th %ile.
...pronounced like the long vowel and first letter of the alphabet...
|
|
|
|
|