iSixSigma

Sample size

Six Sigma – iSixSigma Forums Old Forums General Sample size

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #48816

    Erica
    Participant

    I need to estimate the amount of participants for my study. How do I estimate the standard deviation and the difference in means? Do I choose one or ten articles and take the average?

    0
    #165612

    annon
    Participant

    What are you trying to do?

    0
    #165613

    Widy Dieudone
    Member

    Erica
    I am not understanding you. Give me more to work with.

    0
    #165621

    Six Sigma guy
    Member

    I think you are confusing between sample size estimation for doing hypothesis testing and sample size estimation for doing some study. The difference,standard deviation and power values are required for hypothesis testing. What you need to know for your case is probably confidence levels,confidence interval and margin of error.
    Regards

    0
    #165647

    Erica
    Participant

    I am working with African American Women with intervention of nutrition education and exercise in lowering risk factors for cardiovascular disease. Initially My advisor wanted me to recruit 100 women. One of my committee members wants me to justify the number I need with a formula.

    0
    #165652

    Robert Butler
    Participant

      If this intervention is new and there is no prior data for comparison (in other words data from some other published study which studied AA women with a similar intervention of nutrition, education, and exercise) then you are left with guessing proportional changes. 
      For example, let’s say in the control group 50% of the women have risk factor A and you believe your program will reduce this to 40%.  You can use these two percentages to compute  a sample size.  The drop from 50% to 40% would constitute a 20% decline.   For a 20% drop100 patients (50 control and 50 in new program) would have a power of .17.  You would need a total of 776 patients to get a power of 80%. 
      If you have no prior data then your work would constitute a pilot study and power and sample size calculations aren’t worth much. In those cases when my investigators are up against this situation I usually wind up writing a paragraph to this effect for inclusion in the project proposal and that puts paid to that.
    However:  In Graduate Studies
    Rule #1: Your Advisor is Always Right
    Rule #2: When Your Advisor is Wrong – Refer to Rule #1
    So, let’s assume 100 is all you can afford and you have no prior data and your advisor is demanding high power. In this case you will have to take the percentage prevalence numbers for the controls and cook the percent reduction so that a study of 100 patients will give you a power of 80%.  In the above example this would be a 54% reduction. 
      People do this sort of thing when they are confronted with unreasonable demands and a boss who can’t/won’t listen. However, if there is any chance you have access to a good practicing biostatistician then before doing this I would recommend talking to him/her and asking for help in crafting an answer.

    0
    #165657

    Erica
    Participant

    Thank you Robert. I will find a biostatistician here on campus for help. I will reply with more questions when they come up.

    0
    #165660

    Robert Butler
    Participant

     You are certainly welcome to post questions here, however, if you do have access to a biostatistician I would make it a point to develop a good working relationship with that individual.  Assuming he/she is really interested in doing statistics you should find him/her to be a far better source of information concerning your problem. 
      If you do find such an interested person take the time to run your study protocol past them before starting your work.  He/she should be able to give you suggestions for constructing and running your study that will save you a lot of time and effort. 

    0
Viewing 8 posts - 1 through 8 (of 8 total)

The forum ‘General’ is closed to new topics and replies.