Need a Statistician

[hatman]

One of my college buddies is stuck without a statisticin in the office thismonth, and needs to figure something out that's beyond him. And I can't help him... thought maybe someone here could give him a hand:

I have 2 published studies one of 25 patients that reports 2 adverse reactions and another of 18 patients that reports 4 adverse reactions. Combined that’s 6 adverse out of 53. Crunching % mean and StDev we get 8.0%±5.4%, 22.2%±9.8% and 14.0%±5.3% respectively.<?xml:namespace prefix = o ns = "urn:schemas-microsoft-comfficeffice" /><o></o>
I need to conduct a test with the MINIMUM number of cadaver specimens possible (they are expensive and it's somewhat immoral to waste bodies who are generously donated to science) to prove equivalence. Equivalence is defined as within 1 standard deviation of the existing study. That makes the null-hypothesis "The risk of an adverse reaction with the new system is >=19.3% (14.0+5.3)" within 80% power and 95% confidence. I fully expect to test ~12 specimens and all 12 to have no adverse reaction. I can't find a way calculate a standard deviation of twelve 0's. I'd like the mean of the 12 cadavers to be 80% confident of not falling above the 95% confidence interval of the published studies. I'm not even certain if I'm using the correct language.<o></o>
I know my answer is around 12 but I need to document the decision to use 12 specimens (14 are on order) to have the protocol approved before the lab takes place.<o></o>

 

[mikerickson]

I'm not statistician enough to help, but the requirement "a test with the MINIMUM number of cadaver specimens." reminded me of the (actual) title of a book for Med. students, "Kill as few patients as possible."

 

[sous2817]

I work with 2 statisticians, I'll float it by them and see what they say...

 

[sous2817]

OK, this is what they came back with:

"That makes the null-hypothesis 'The risk of an adverse reaction with the new system is >=19.3% (14.0+5.3)' " - They said that this is the 'Alternative Hypothesis', not the 'Null Hypothesis'. To calculate the sample size, it is required to know the null hypothesis, the proportion would be if the medicine works good"

Not sure if that helps, but that's what they came back with.

 

[Gerald Higgins]

Would statisticians' cadavers be any use to you ?

 

[hatman]

Thanx for looking at this for me. I'll pass it along to my friend, though I'm not sure how much it'll help.

Any reason why I wouldn't get the automated E-Mail notification for replies to this thread? I'm subscribed to it, and MrExcel isn;t on the blocked list (that I know of)... I still get notifications from the Invitation to become a member of the DRAFT....