Question on power, precision, and sample size

[Matt Price]

Hi there,

I have a specific power calculation that I would like to make, but I can’t figure out how to write the code for this.

We have a cohort of 400 persons, in two equal age strata (200 adolescents aged 15-19 and 200 young adults aged 20-24). Participants will be followed for two years. I would like to know our power to detect a difference in HIV incidence, and graph it for the following:

We assume our HIV incidence among the young adults is 3% per year, for two years.

What is our power to detect a difference in the younger cohort, if the HIV incidence in adolescents is 4%, 6%, 8%, or 10% per year, for two years? I’d like to graph the figure for power by incidence.

Can Stata graph two lines, one showing the comparison if young adults observe a 3% incidence per year, for two years and one if young adults observe a 2% incidence per year, for two years.

Any help you could provide would be very much appreciated.

[Bruce Weaver]

What kind of model do you plan on using? Some kind of count regression model, perhaps, so that Exp(B) = an incidence rate ratio? Thanks for clarifying.

[Matt Price]

Hi Bruce, I had anticipated keeping it as simple as possible, so something like a regression model would be fine. In the past, I have used the Stata commands for power calculations - but I don't require them specifying an underlying model (this may however belie my knowledge of how these commands work!)