If I don't have the data, but only have the estimates of mean and variance of two independent Gamma distributions. What type of test I can use to test the null hypothesis μ1=μ2?
Asked
Active
Viewed 611 times
1

kjetil b halvorsen
- 63,378
- 26
- 142
- 467

Q_Li
- 377
- 2
- 12
-
1How were the estimates obtained? Do you know the sample size for each? Do you know the the shape parameter of these gammas or not? – Glen_b Apr 15 '14 at 06:18
-
2The [help](http://stats.stackexchange.com/help/on-topic) says "*cross-posting is not encouraged on SE sites. Choose one best location to post your question. Later, if it proves better suited on another site, it can be migrated.*" – Glen_b Apr 15 '14 at 07:13
-
I treated my data as Negative Binomial distribution. And I know the Negative Binomial distribution is a mixture of Poisson and Gamma distribution. So I estimated the variance and mean of the Gamma distribution, this post (http://math.stackexchange.com/questions/752964/can-i-estimate-variance-of-gamma-from-negative-binomial-distributed-data-given) explains how I estimated the variance and mean of the Gamma distribution. Sample size of Negative binomial is small, only 5. – Q_Li Apr 18 '14 at 23:55
-
Have you considered Bayesian methods? Given what you lay out in your previous posting Bayesian methods make more sense than frequentist approaches. With that said, fancy mathematics is not a cure for micronumerosity. What is your estimate of p in the Gamma prior? – Dennis Jun 13 '14 at 04:07
-
See https://stats.stackexchange.com/questions/48378/difference-of-gamma-random-variables Maybe it can help – kjetil b halvorsen Sep 12 '18 at 12:26