Estimating population mu from multiple small samples

#1
Hi,

I am trying to consider the best option for the following situation I’m dealing with.

I have 3 sets of results from ‘skid test’ accelerometer data taken by 3 different people, sample sizes 3, 3 and 4. The tests were conducted under the same conditions (car and road) and each one provided an average friction coefficient for that test. I am trying to find the best estimate of the actual (population) friction coefficient.

Would it be most accurate to calculate this using a single sample of size n=10, or by finding the mean of the three sample means? I have calculated the mean, s, and confidence interval using a student t table for each of the three samples from each person, but how do I then apply this to the best estimate of the ‘actual’ friction coefficient?

Thanks
F