0
$\begingroup$

I am a doctor and right now I am doing medical research. I have some questions about cohen's d. I try to research it on the internet but it seems there is not much answer to this question.

My project have non normal distributed data. I want to compare variables using independent t-test, therefor I log transformed my data. Another part of analysis is effect size (Cohen's d). I am wondering should I calculate Cohen's d from raw data, log transformed data, or back transformed of log data. Since Cohen's d calculation involves sd so I am not sure can I back transform SD and use it.

Ps. Further question is very basic but I just want to make sure because I am an doctor not an expert in statistic. Please help me confirm that I don't get it wrong.

I log transformed my data. Calculating it using independent t-test. The result are mean of log data of 2 groups. I back transform this mean of log by 10 power (mean of log) and present it in my paper as geometric mean. It seem like presenting geometric mean with SD or SE is not appropriated. Should I present geometric mean with 95% CI instead? I calculate 95% CI in SPSS using one sample t-test and back transform upper and lower 95% CI using 10 power that number.

Am I getting it right?

Improve this question
$\endgroup$
4
  • $\begingroup$ It will be helpful if you can give some details about your data. What is the sample size? Are the values all positive or negative also? It may be best if you can post a sample of your data in your question. $\endgroup$
    – rnso
    May 10, 2015 at 9:44
  • $\begingroup$ There are overall 1500 subjects. One group contains 1300 subjects and another group contains 150 subjects. All the values are positive. It is cross sectional study. $\endgroup$
    – tomus riddleburg
    May 10, 2015 at 13:37
  • $\begingroup$ For this sample size, unpaired t-test may not be bad. See this page: jerrydallal.com/LHSP/STUDENT.HTM (3rd last paragraph). $\endgroup$
    – rnso
    May 10, 2015 at 14:47
  • $\begingroup$ Thanks for the comment. What about Cohen's d? Should I calculate it from raw data, log transformed data, or back transformed of log data. $\endgroup$
    – tomus riddleburg
    May 10, 2015 at 15:40

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.