reliable change index and reliability coefficient

#1
Hi. I'm a bit embarrassed at my inability to figure this out even with the vast amounts of information available to me via the internet. Any help is appreciated. (Feel free to make fun of me too.)

I am trying to calculate RCI (reliable change index) which I am able to do with a simple formula - SEM*(standard deviation) X SQRT(1-reliability) I have everything I need for this calculation except a reliability number. It's making me crazy.

All I'm trying to do is come up with a reliability coefficient for the 34 client scores below. As I understand it, the reliability should be pretty close to the average (which for the scores below is 83). But using the formulas I'm finding, I keep coming up with weird numbers like 20. Any help at all would be greatly appreciated! Thanks in advance.

Client Score
A 83
B 80
C 80
D 83
E 86
F 80
G 85
H 79
I 80
J 89
K 87
L 90
M 86
N 74
O 86
P 80
Q 73
R 81
S 92
T 87
U 85
V 89
W 84
X 81
Y 78
Z 86
AA 79
BB 86
CC 84
DD 79
EE 81
FF 90
GG 85
HH 89
 

CB

Super Moderator
#2
You can't really estimate reliability simply from a set of scores without any other information. Do you have access to the responses to the actual individual items used to produce these scores? If so you could calculate a reliability estimate such as Cronbach's alpha, McDonald's omega, Raykov's rho, etc.

You could also obtain a reliability estimate from a previous study using the instrument, although there's no guarantee that the reliability of observations found in another sample will also apply to your sample.

Do make sure you know what is meant by reliability... it's not an average. Some background reading might both help you understand how to do the calculations better, but also to decide whether RCI is really telling you what you want to know. A good intro to psychometric theory is Raykov & Marcoulides, 2011.
 

Lazar

Phineas Packard
#3
Surely you also need two time points (i.e. two vectores of client scores). With only one measure the idea of reliable change is meaningless right?

P.S. For RCI I believe you can use the test-retest correlation for reliability if memory serves me correct.
 

Karabiner

TS Contributor
#4
All I'm trying to do is come up with a reliability coefficient for the 34 client scores below.
What are these scores? What was measured, and
where does your measurement instrument come from?

With kind regards

K.