Hi,

Lately I have been studying about different statistical methods that can be used to study significant differences among multiple groups. I understand that multiple t-tests are not ideal because of the increasing type 1 error. I read about ANOVA and I understand now that its result depends on the variance between the groups and the variance among the groups.

I am currently designing a study protocol in which we have 1 control group and 3 experimental groups (treatment groups). In short, we will transplant cardiomyocytes (with or without other cells) to an animal model and explant later. At the time of explantation, we will study several histological and immunohistocemical parameters quantitatively.

- In the control group (CG), we transplant cardiomyocytes only.

- In experimental group 1 (EG1), we transplant a mix of cardiomyocytes and fibroblasts.

- In experimental group 2 (EG2), we transplant a mix of cardiomyocytes and peripheral-blood mononuclear cells.

- In experimental group 3 (EG3), we transplant a mix of cardiomyocytes and mesenchymal stem cells.

For all our results, the control group will be our baseline and we expect (based on a preliminary study) that the results in EG1, 2 and 3 will be higher compared to control. More specifically, we expect that EG3 will give the best results and then EG2, EG1 and finally control: EG3 > EG2 > EG1 > CG.

Most importantly for us is:

- does EG1 give a better result than control?

- does EG2 give a better result than control?

- does EG3 give a better result than control?

Secondly, we would like to know whether the 3 experimental groups are different from each other.

My question is whether ANOVA is the method of choice here to study the null hypothesis that the means are equal. Before I have seen ANOVA tests in independent groups, without a control group. Here, I have a control group, so I assume that the data from the experimental groups should be mainly compared to the control group and not to the mean of all groups.

In other words, our “fear” is that results from EG1 and EG2 bridge the gap between the control group and EG3, and therefore would make the whole observation non significantly different. I suspect that EG3 would be different from the control using a t-test, because it does not take into account the 2 other groups. With ANOVA the mean of all the samples will change and bridge the difference between the control and EG3.

I am getting mixed opinions from colleagues, so if anyone could share any insights on this, I would be very grateful. Thanks

More Linus Van Marsenille's questions See All
Similar questions and discussions