m********9
发帖数: 208 | 1 I have a design to compare the difference between Method 1 and Method 2 (for
short, M1 and M2). In this design, I have two sections (S1 and S2), each
have 20 observations, and a chosen topic with four levels (A, B, C, and D)
The design is as follows.
A B C D
_____________________________
S1 | M1 M2 M1 M2
|
S2 | M2 M1 M2 M1
Specifically, for topic A in S1, we use M1 method, for topic A in S2, we use
M2 method, etc. I want to test the significance between M1 and M2. I may
consider Sections as between-subject effect, and Methods as within-subject
effect, but I am not sure how to treat the chosen topic (between-subject,
within-subject, or something else?). Any input about analyzing this design
will be greatly appreciated. Thanks. | u*h
发帖数: 397 | 2 why not full factorial design?
factor 1: ABCD (4 levels)
factor 2: S1, S2 (2 levels)
factor 3: M1, M2 (2 levels)
full factorial design: 4*2*2 = 16 levels.
each level can have 2 repeats, and total is only 32.
You want to design experiment or you just need analysis the result? | m********9
发帖数: 208 | 3 I just want to analyze the data. Each level should have 16 repeats, and
total is only 16*16. For S1 with topic A and M1, I have 16 observations. | u*h
发帖数: 397 | 4 what is your assumption? like this?
observation value =
effect of (A or B or C or D) +
effect of (S1 or S2) +
effect of (M1 or M2) +
error
if this is your model, one way I can think is to use linear regression,
create 3 dummy variables for ABCD
1 dummy for S1/S2
1 dummy for M1/M2
test the last dummy variable coefficient not equal to 0 | m********9
发帖数: 208 | 5 Thanks. That makes lots of sense. |
|
