Hello,

Since you are testing both between groups and with group significance. I recommend using an ANOVA. I do not think Cohen's D can be used as a measure here as it just provides the scale/size of the effect depending on the t statistic. Create two columns in your data set: one that specifies the population group and one that distinguishes between control and experimental. You can use the F t-test to compare between populations and then Tukey's to find within group differences. Also look at the partial eta squared values.

Hopefully this helps you.

Regards,

Hi Vishal,

Thank you for your response. I forgot to mention that for one population I only have the mean, sd and sample size.

Would you think that in that case an anova still works?

What I did now is use a t test, which gives me the results I would expect, but my data is not nornormally distributed so I doubted whether this was the good way to go. Thank you.

Best regards,

Aloys

Hi,

So, you don't have any data for one population except for the summary values of mean, SD, and sample size?

I have the data of one population but only the three statistics of the others. I do it for ccomparing my results with the results of another study.

Hello,

If the three statistics are for the difference between the control and experimental group, then you can find the difference between your experimental and control group and then carryout a t-test and f-test. However, if you only have the three statistics for the entire population without them being for control and experimental, I do not think there will be a logical comparison. You will only be able to compare if the two populations are different and not the within group differences (control and experimental).

The normality should not be a major issue as the samples are independent (both populations are different). If i can get a better look at the data, I can provide a more adequate approach.

Thanks

Hi,

For one population I have the entire data and for the other I only have the three statistics. For the population I do not have the entire of I do have the seperate three statistics of the control group and the experiment group.

I added an overview of the statistics of both populations. Thank you!

Hi,

Ah I see. The importance of having the complete data is that you can plot and check for the linearity, randomness, and normal distribution. Now since you do not have the data of one population, you will never know if linearity was satisfied (and this is very important). You can make an assumption and move forward but I do not know how valid the results will be.

And like you show in the summary table, I think you can carry out another t-test on the difference of control and experimental for the two populations. I think this is the best you can do with the data because you do not have the sum of residuals or mean square thus an F-test/chi-squared test might not be possible.

Most people do not question linearity and just focus on p-value, so you might get away with it. :)

Hope this helps you!

Best,

Vishal