I have a question regarding the significant difference testing.

I have a question regarding the significant difference testing. Can any expert guild me here.

I have an average purchase amount about $102 in the population(sample size 8,091). The standard deviation is about $189(big). The whole data heavily right skewed. I want to take out the outliers when I do statistical difference testing. I have two questions here.

1. Can I get it directly from the data or I need to do LOG transformation? Since the data looks normal after the log transformation.
2. How can I take off the outliers? From the original data or Log transformed data. Should I use 3xstandard deviation rule or top 5% extreme value rule?

Any help will be highly appreciated.

所有跟帖： 

• If you have a large sample, you can do whatever way. -Pink俠- ♀ (41 bytes) () 12/12/2012 postreply 18:32:51