Independence of the data

[Vahid]

Hello.
I assigned 18 participants to the experimental group (n=10) and control group (n=8). The experimental group received intervention A and the control group received intervention B. After 6 weeks, I applied intervention A to the control group (n=8). Then, I mixed the data of both groups which received intervention A (18 participants) and compared them with the data of the control group which received intervention B (n=8). The results are very interesting! I received a comment from a researcher that it is a violation of the dependence of data. Would you please assist me to find support for the way that I mixed the data? In fact, I need to support the way that I mixed the data to publish the data.

I appreciate your kind attention to my request.

 

[BigBeachBanana]

Then, I mixed the data of both groups which received intervention A (18 participants) and compared them with the data of the control group which received intervention B (n=8).

I don't understand which groups you are mixing.

 

[Vahid]

I don't understand which groups you are mixing.

Thank you very much for the reply.
The experimental group received only one intervention; intervention A.
The control group received two interventions: first intervention B and after six weeks they received intervention A (the same intervention that was applied to the experimental group).
I mixed the post-intervention data of the experimental group which was collected after intervention A with the post-intervention data of the control group after applying the intervention (n=8) A together (n=10+8=18) and compared them with the post-intervention data of the control group which was collected after applying the intervention (n=8) B.

 

[BigBeachBanana]

Thank you very much for the reply.
The experimental group received only one intervention; intervention A.
The control group received two interventions: first intervention B and after six weeks they received intervention A (the same intervention that was applied to the experimental group).
I mixed the post-intervention data of the experimental group which was collected after intervention A with the post-intervention data of the control group after applying the intervention (n=8) A together (n=10+8=18) and compared them with the post-intervention data of the control group which was collected after applying the intervention (n=8) B.

Effectively, what you did was conduct research on two groups:
Group 1: Intervention A (n=10) and
Group 2: Interventions A and B (n=8)
You don't have a control group with intervention B only.
UNLESS you can justify that after six weeks, intervention B's effect has completely worn off and has no effect on the outcome of intervention A.
That's a BIG IF.
You said you saw an interesting result in group 2, but that might be because of their exposure to intervention B already.
Typically, researchers don't do this because of biases. Since group 2 has already been exposed to intervention B, which might provoke group 2 to behave a certain way. If you apply intervention A on top on their prior experience with intervention B, they might've acted differently compared to a group who experienced either intervention for the first time.

In short, I agree with the comment the researcher made.

 

[Vahid]

I gratefully appreciate your kind attention to my question.

Applying nonparametric analysis, I found out that there was not any difference before intervention A in both groups (Mann-Whitney U p=.184). Additionally, there was not any difference before intervention B and before intervention A as well as after intervention B and before intervention A in the control group (Wilcoxon p≥.726). On the other hand, the comparisons were done using nonparametric tests which do not rely on the assumption of independence of data. Altogether, I think it is possible to trust the interesting findings of the intervention.