L**y
发帖数: 128 | 1 就是拿一个measure的得分去跟另一个measure比较的，
I am stuck with the statistic part, need help desperately. :( | r******s
发帖数: 2155 | 2 what's your question?
【在 L**y 的大作中提到】
: 就是拿一个measure的得分去跟另一个measure比较的，
: I am stuck with the statistic part, need help desperately. :(
| L**y
发帖数: 128 | 3 I am comparing a new assessment tool with the already validated tool which, in
my case, is BDI-2.
Since I only care about whether the new tool are concurrently valid in
measuring child's cognitive ability, I only select about 20 items from the
whole battery. I do the same for BDI-2 as well (the whole BDI-2 is divided
into 6 domains, consist of 100 items. But I am only going to pick about 20
items that measure child's cognitive ability).
For items in both of my measures, there are 3 possible resp
【在 r******s 的大作中提到】
: what's your question?
| r******s
发帖数: 2155 | 4 I'm sure that ARIMA would be able to answer your question better.
I can only ask you several questions and hopefully help you in the right
direction.
First, it sounds like your measure is multiple choice. Then you cannot use
pearson
correlation because it is for continuous variables. You can check Nunally and
Berinstein's
measurement book (94) or Lord and Novick's (1968) classic book.
Second, I'm not familar with Rasch modeling. But if I were you, I'd do two
things.
1. Check what exact the model
【在 L**y 的大作中提到】
: I am comparing a new assessment tool with the already validated tool which, in
: my case, is BDI-2.
: Since I only care about whether the new tool are concurrently valid in
: measuring child's cognitive ability, I only select about 20 items from the
: whole battery. I do the same for BDI-2 as well (the whole BDI-2 is divided
: into 6 domains, consist of 100 items. But I am only going to pick about 20
: items that measure child's cognitive ability).
: For items in both of my measures, there are 3 possible resp
| r******s
发帖数: 2155 | 5 BTW, I'm not sure what you are doing is concurrent validity.
If your new assessment tool is a test of cognitive validity, which is same as
BDI-2,
then it is convergent validity.
in
2.
items
me
【在 L**y 的大作中提到】
: I am comparing a new assessment tool with the already validated tool which, in
: my case, is BDI-2.
: Since I only care about whether the new tool are concurrently valid in
: measuring child's cognitive ability, I only select about 20 items from the
: whole battery. I do the same for BDI-2 as well (the whole BDI-2 is divided
: into 6 domains, consist of 100 items. But I am only going to pick about 20
: items that measure child's cognitive ability).
: For items in both of my measures, there are 3 possible resp
| A***A
发帖数: 98 | 6 Loly,
First, just as restless noted, the observed response 0-1-2 is categorical, so
Pearson's correlation, which is for continuous outcomes, is probably not
adequate.
Second, since your instrument is new, item response theory models, with the
Rasch model as a very special case, provide powerful tools for you to conduct
item analyses in ways that are hard to imagine with classical test theory.
Furthermore, fitting an item response model (note: NOT the Rasch model)
enables you to obtain more effi | L**y
发帖数: 128 | 7 太感谢了！！！
谢谢restless as well.
让我先把基础知识补起来，再来向大家请教探讨。
so
conduct
that
provide
theory
data.
that
responses
able
only
continuous
【在 A***A 的大作中提到】
: Loly,
: First, just as restless noted, the observed response 0-1-2 is categorical, so
: Pearson's correlation, which is for continuous outcomes, is probably not
: adequate.
: Second, since your instrument is new, item response theory models, with the
: Rasch model as a very special case, provide powerful tools for you to conduct
: item analyses in ways that are hard to imagine with classical test theory.
: Furthermore, fitting an item response model (note: NOT the Rasch model)
: enables you to obtain more effi
| L**y
发帖数: 128 | 8 Million thanks.
我的相关committee member 要下个月才有空和我讨论这个问题，我是等不及了，先自学
一下，到时候不至于显得很stupid.
and
Since
which,
and
am
here.
I
【在 r******s 的大作中提到】
: I'm sure that ARIMA would be able to answer your question better.
: I can only ask you several questions and hopefully help you in the right
: direction.
: First, it sounds like your measure is multiple choice. Then you cannot use
: pearson
: correlation because it is for continuous variables. You can check Nunally and
: Berinstein's
: measurement book (94) or Lord and Novick's (1968) classic book.
: Second, I'm not familar with Rasch modeling. But if I were you, I'd do two
: things.
|
|
