Hello, after a unsuccessful search in lists maliarchive I wonder how I could estimate the power of a sample size related to an unknown population. Given the following (fake))situation: I do have a database containing about 5 millions observations over 70 variables. I would like to compute (as epidemiologists are used to) the required size of a sample to do some test on a test sample (test data), later doing some subsequent analysis of a new sample to build a prediction model. Help facilities of R show some entries regarding power, but none of them seemed to be appropriate for my purpose (maybe I am wrong, but sometimes I have some difficulties to decipher the message of those tiny hints for available packages) I would appreciate somebody effort to point me to the right package/function to archieve this task! regards Thomas
Power sampling
3 messages · Uwe Ligges, Thomas Friedrichsmeier
Thomas Sch??nhoff wrote:
Hello, after a unsuccessful search in lists maliarchive I wonder how I could estimate the power of a sample size related to an unknown population. Given the following (fake))situation: I do have a database containing about 5 millions observations over 70 variables. I would like to compute (as epidemiologists are used to) the required size of a sample to do some test on a test sample (test data), later doing some subsequent analysis of a new sample to build a prediction model. Help facilities of R show some entries regarding power, but none of them seemed to be appropriate for my purpose (maybe I am wrong, but sometimes I have some difficulties to decipher the message of those tiny hints for available packages)
You have to tell us for which test you are going to calculate the power ... (and there might be nothing, since calculating the power precisely is not always that easy). Uwe Ligges
I would appreciate somebody effort to point me to the right package/function to archieve this task! regards Thomas
______________________________________________ R-help at stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Hello Uwe, Uwe Ligges schrieb:
Thomas Sch??nhoff wrote:
You have to tell us for which test you are going to calculate the power ... (and there might be nothing, since calculating the power precisely is not always that easy).
Given my example from the first message I asked for a function which enables me to calculate a reasonable sample size. I don't know the true mean or standard deviation of the population, I only know: n= 5.000.000 observations over 70 variables determined alpha = 0.01 I want to know how to generate a sufficiently sized sample based upon above mentioned facts to make some valid predictions regarding my population. All I can hink of for now is that a two-tailed power test is required to find out if H0= random effect or H1= no random effect hypothesis is accepted/rejected. In epidemiological studies this situation is described like this: How many cases do I have to include in my sample (s) to gain some representative results from a unknown population (=true mean, std-deviation etc.)? How can I approach a situation like this in R ? Regards Thomas platform i386-pc-linux-gnu arch i386 os linux-gnu system i386, linux-gnu status major 2 minor 0.1 year 2004 month 11 day 15 language R