Calculating Sample Size for an Olympic Sprinter Experiment

[Pistone]

Dear forum pal,

Here is my problem:

- We would like to buid an experiment in order to check the effect of a new energy drink that is destined to the olympic sprinters (100m).

- Here are our requirements for the experiment.

- Type I error = alpha = less than 0.05
- Power = 1-beta = better than 0.80

If the results difference over 100 meter between someone who used the drink and someone who did not is 0.1%, what is the required sampling size ?

Same question for a difference of 0.2 %, 2.5%, 3.6%, 4.0% and 4.7%


Please try to help us.
If our question lacks any other information please let me know.

Pistone.

 

[EnumaElish]

Think of this as the mirror image of the "usual" problem "given sample size, the two means and the two standard dev.s (under the null and the alternative), determine the alpha and the beta."

 

[Pistone]

Thank you EnumaElish

Dear EnumaElish,

Thank you very much for taking the time to answer me.

Unfortunately for me, I am very weak in statistics as my basic expertise is Food Formulation and Innovation.

Please can you help me more ?

Thank you again for your kind help.

Best Regards,
Pistone.

 

[Pistone]

Additional Information

Dear Forum Pals,

It looks like my request lacks some basic information.

Please receive the missing information:

- The null hypothesis is that there is no difference between a sprinter that have consumed the drink and a sprinter that have consumed mineral water.

-The Alternate hypothesis is that there is a difference between the 2 sprinters and that this difference improves the results in seconds (over 100m) of 0.1% (sprinter with energy drink better than sprinter with water). Other experiments would have alternate hypothesis of improving the results for 0.2%, 2.5%, 3.6%, 4.0% and 4.7%.

-Concerning the beta, since beta=(1-statistical power of the experiment) and since such experiments in order to be valid should have a power of at least 0.80 then beta=0.2

I have found a nice site that makes automatic calculations I don't know how to use it properly (I am weak in statistics, my basic expertise is Food Formulation and Innovation). I wonder if we can post links on this forum but you can find the tool on dssresearch.com then click researcher's toolkit and then sample size calculator and then percentages. But you are the experts, You surely know better than me how to find the answer...

Thank you again for your precious help, it is highly appreciated.

Best Regards,
Pistone

 

[EnumaElish]

Yes you can post links. See http://en.wikipedia.org/wiki/Sample_size#Required_sample_sizes_for_hypothesis_tests for an explanation of the problem and the solution method.

 

[almaldon]

Hello all,
I am working with molecular data and my aim is determinate if there are differences in the quantity of 17 molecular markers between 4 populations (N=72 [rows] x 17 markers [columns] each one found in different proportion in the samples, then I calculate the mean by population by each marker and try to make statistics). How can I choose a good statistical test with these sample size? How much affect the sample size the statistical test? The data not follow a normal distribution and I use Kruskal-Wallis test to make the comparison, but I am not sure if this test suits with this kind of data. Please, help!. Thank you :)

 

[EnumaElish]

K-W works as a good approximation, with only constraint being that the K-W test "assumes an identically-shaped distribution for each group"; see http://en.wikipedia.org/wiki/Kruskal-Wallis_test

 

[almaldon]

Thanks EnumaElish, I taken into account this issue.
I have another question... How the sample size affect the selection of the statistical analysis tools? Any reference for that?
Again thank you! :)