How are your groups structured? You mention sham-control and then two different montages. What does this look like for within and between groups allocations? We need a lot more description to provide a thorough answer. Your test design is also going to depend on sample size, whether or not there are any relationships between the variables, and how interested you are in the factor of time (e.g. do you want to measure the line/curve, or are you just interested in the change relative to baseline). My very vague suggestion would be to look into linear mixed models if you want to make the best use of your time variable. Otherwise, correct your outcome measures relative to baseline and then use Multivariate ANOVA. Try to follow up with a departmental statistician if you can.

Hi Nik Murphy

Thank you for your meticulous reply!

35 participants (stroke patients with unilateral impaired upper limb) were randomly assigned to one of three groups: Group A (n= 11) received tDCS with a specific montage to target a specific cortical area, Group B (n= 13) received tDCS of different montage to target another cortical area, Group C (n= 11) received sham stimulation. All the participants received 10 sessions over 2 consecutive weeks (5 session/week). The objectives of the study are:

1) To investigate and compare the effects of two tDCS montages at different cortical sites by measuring the behavioral/clinical outcomes of the impaired upper limb (T1-T0)* .

2) To investigate the long-term effects of tDCS on upper limb motor performance (T2- T1)*.

N.B. 4 outcomes were used to measure the motor ability and dexterity. We are not interested in the relationship between the variables but we need to determine which outcome showed better treatment response and better long-term effect ).

* The time points are defined as follows: baseline measurements (T0), post-stimulation measurements immediately after the 10th session (T1), and long-term effects 4 weeks after the end of the stimulation period (T2).

Thanks. In this case I would use linear mixed modeling to analyse your data.

Nik Murphy

Why not the repeated measures ANOVA or MANOVA?

Many thanks

If you are interested in using the full breadth of your effect of time then repeated measures ANOVA is not substantially powered.

I'm not interested to treat the time as continuous, it is basically pre-post design with three repeats.

Based on the design you've provided, it sounds like a suitable statistical test for this study would be a mixed-design ANOVA. This test would allow you to compare the effect of the two tDCS montages at different cortical sites (Group A and Group B) on behavioral and clinical outcomes of the impaired upper limb (measured at T2, T1, and T0) while controlling for the sham stimulation group (Group C). This test allows you to compare the effect of the two tDCS montages across different groups (between-subject factor) and also across different time points (within-subject factor).