Dear all,
Given that I have a carryover effect in some of my seeds - I decided to add a 2nd-level covariate that accounts for order: it codes 1 for each participant that had the real treatment administered first, and 0 otherwise.
Is this considered as a viable approach for controlling for order in such an experiment?
Thanks in advance,
Yuval.
[i]Originally posted by Yuval Argaman:[/i][quote]Dear Alfonso and CONN community,
I'm analyzing the resting-state data from my study, in which each participant underwent 2 scans X 2 conditions.
This means that I my conditions are: Pre_Real, Pre_Sham, Post_Real, and Post_Sham.
Accordingly, I defined my repeated-measures ANOVA: AllSubjects [1], and Pre_Real, Pre_Sham, Post_Real, Post_Sham [-1 1 1 -1] (as specified in the CONN manual, p.26).
However, I have a reason to suspect that there might be a carryover effect. Therefore, I decided to perform a paired t-test, comparing only the Pre_Real and Pre_Sham, so I selected only the 2 of them and defined the contrast as [-1 1]. Here, I got some significant results (which is a cause for concern)
Additionally, I have a reason to believe that controlling for the order in which the participants received their treatments - would neutralize the order effect.
To this end, I defined the order effect using 2 second-level dichotomous covariates - Real_first and Sham_first, which are the opposite of each other. I'm not sure that this is the way to do it, though.
Given the experimental design's within-subject nature, I'd like to ask the following questions:
1. Did I define the carryover effect test correctly? or should I instead use the contrast Pre_Real, Pre_Sham, Post_Real, Post_Sham [-1 1 0 0]?
2. How should I define the order effect in order to check for its influence and possible association with the carryover effect?
Sorry for the long post and thanks in advance,
Subtly,
Yuval.[/quote]
Given that I have a carryover effect in some of my seeds - I decided to add a 2nd-level covariate that accounts for order: it codes 1 for each participant that had the real treatment administered first, and 0 otherwise.
Is this considered as a viable approach for controlling for order in such an experiment?
Thanks in advance,
Yuval.
[i]Originally posted by Yuval Argaman:[/i][quote]Dear Alfonso and CONN community,
I'm analyzing the resting-state data from my study, in which each participant underwent 2 scans X 2 conditions.
This means that I my conditions are: Pre_Real, Pre_Sham, Post_Real, and Post_Sham.
Accordingly, I defined my repeated-measures ANOVA: AllSubjects [1], and Pre_Real, Pre_Sham, Post_Real, Post_Sham [-1 1 1 -1] (as specified in the CONN manual, p.26).
However, I have a reason to suspect that there might be a carryover effect. Therefore, I decided to perform a paired t-test, comparing only the Pre_Real and Pre_Sham, so I selected only the 2 of them and defined the contrast as [-1 1]. Here, I got some significant results (which is a cause for concern)
Additionally, I have a reason to believe that controlling for the order in which the participants received their treatments - would neutralize the order effect.
To this end, I defined the order effect using 2 second-level dichotomous covariates - Real_first and Sham_first, which are the opposite of each other. I'm not sure that this is the way to do it, though.
Given the experimental design's within-subject nature, I'd like to ask the following questions:
1. Did I define the carryover effect test correctly? or should I instead use the contrast Pre_Real, Pre_Sham, Post_Real, Post_Sham [-1 1 0 0]?
2. How should I define the order effect in order to check for its influence and possible association with the carryover effect?
Sorry for the long post and thanks in advance,
Subtly,
Yuval.[/quote]