Should the number of trials be balanced for each condition within a contrast?

1

Hi Experts!

I have a statistical question. I will be carrying out a first level analysis on a set of task-fMRI data with two conditions, i.e. condition high - with stimulus that appear 70% (420 trials across 4 blocks) of the time, and condition low - with stimulus that appear 30% (180 trials across 4 blocks) of the time.

The event-related trial sequence is designed with genetic algorithm (Wager and Nichols, 2003) to maximise statistical power.

Given the nature of the the two conditions, the high condition will have alot more trials than the low condition. Is it necessary to balance the number of trials between the two conditions in the analysis?

This could be done by randomly sampling and discarding trials from the high condition. However, this means I will loose more than half of the trials in the high condition.

What are your thoughts on this?

Thank you!

2

all of those blocks are in the same run?

if so I would not balance

3

Agree with Remi-Gau. The unbalance on number of trials does not bias parameter estimates AFAIK.

4

Thanks for the response!!

opps, sorry “it is 70% (420 trials across 4 RUNs) of the time, and condition low - with stimulus that appear 30% (180 trials across 4 RUNs) of the time.”

Additional info:

There are 12 blocks across 4 runs.

this is a serial reaction time task, there are 4 fingers that maps to 4 stimulus which yields 4 types of unique sequences. In each sequence, one finger will correspond to a high probability stimulus. Each of this sequence is repeated 3 times which yields 12 blocks (3repeats x 4 sequences) across 4 runs.