w********u 发帖数: 328 | 1 the sample size is extremely small but the population is very large, for
example, the frequency of some cancer is extremely low.so it doesn't make
sense to say that the sample follows normal dist.
my question is
1 at least how many sample roughly i need to conduct a power analysis or
what's the lowest frequency to conduct a power analysis
2 the proc power in SAS is for normal dist.power analysis, so which method
could be used in this case, SAS preferred.
Thank you! |
D******n 发帖数: 2836 | 2 ur understanding of some statistical issues is wrong....for example, sample,
population, ....
【在 w********u 的大作中提到】 : the sample size is extremely small but the population is very large, for : example, the frequency of some cancer is extremely low.so it doesn't make : sense to say that the sample follows normal dist. : my question is : 1 at least how many sample roughly i need to conduct a power analysis or : what's the lowest frequency to conduct a power analysis : 2 the proc power in SAS is for normal dist.power analysis, so which method : could be used in this case, SAS preferred. : Thank you!
|
h****s 发帖数: 16779 | 3 What test will you use?
【在 w********u 的大作中提到】 : the sample size is extremely small but the population is very large, for : example, the frequency of some cancer is extremely low.so it doesn't make : sense to say that the sample follows normal dist. : my question is : 1 at least how many sample roughly i need to conduct a power analysis or : what's the lowest frequency to conduct a power analysis : 2 the proc power in SAS is for normal dist.power analysis, so which method : could be used in this case, SAS preferred. : Thank you!
|
w********u 发帖数: 328 | 4 so would you please correct my wrong understanding? |
w********u 发帖数: 328 | 5 I think in this case, the only method i could choose is nonparametric, but i
don't know exactly how, could you give me some hint to the code in SAS?
THANK YOU. |
l*********s 发帖数: 5409 | 6 Rare discrete events can be assumed to follow Poisson distributions. |
s*r 发帖数: 2757 | |
w********u 发帖数: 328 | 8 but i'm pretty sure it can't follow any particular distribution, it's a very
practical problem |
h****s 发帖数: 16779 | 9 Still do not understand what you are going to do.
Maybe you can give a makeup data example about the data structure and what
you want to test. You may do power analysis for this data first. And the
sample size calculation will easily follow.
The power and sample size analysis can be carried out only after the test
has been decided. Simply saying "need to conduct a power analysis" without
what kind of test makes no sense. |
w********u 发帖数: 328 | 10 I want to calculate the POWER.
SAS code:
proc power;
onesamplemeans
mean = 8
ntotal = 150
stddev = 40
power = .;
run;
This method makes sense when it follows normal dist., but my case is that
the sample size is extremely small but the population is very large, i can't
say that it follows normal dist., i assume it doesn't follow any particular
dist. but i haven't got the data yet, i wonder which method could be used
in this case and what's the sample size needed to conduct any analysis. for
examp |
w********u 发帖数: 328 | |
A*******s 发帖数: 3942 | 12 describe a little bit about your data.
【在 w********u 的大作中提到】 : the sample size is extremely small but the population is very large, for : example, the frequency of some cancer is extremely low.so it doesn't make : sense to say that the sample follows normal dist. : my question is : 1 at least how many sample roughly i need to conduct a power analysis or : what's the lowest frequency to conduct a power analysis : 2 the proc power in SAS is for normal dist.power analysis, so which method : could be used in this case, SAS preferred. : Thank you!
|
h****s 发帖数: 16779 | 13 This code is for one-sample t-test.
Nonparametric generalization of the one-sample t-test is the Wilcoxon signed
rank test.
You can check out this paper for the power and sample size calculation:
Gwowen Shieh; Show-Li Jan;Ronald H. Randles (2007). Power and sample size
determinations for the Wilcoxon signed-rank test. Journal of Statistical
Computation and Simulation, Volume 77, Issue 8, pages 717 - 724.
StudySize 2.0 may realize some calculation.
【在 w********u 的大作中提到】 : I want to calculate the POWER. : SAS code: : proc power; : onesamplemeans : mean = 8 : ntotal = 150 : stddev = 40 : power = .; : run; : This method makes sense when it follows normal dist., but my case is that
|