help to plot 3 groups in 5 different conditions.

24 views
Skip to first unread message

Eduardo Umeoka

unread,
Aug 6, 2020, 10:14:39 AM8/6/20
to estimationstats
Dear developers and users.
I would like to plot a data set where I have 3 groups (control, disease 1, disease 2) in 5 different conditions (placebo, drug A, drug B, C, D and E).

When I did 2-way ANOVA I could make compassions between groups for each treatment. Ex. control vs disease 1 vs disease 2 in Placebo, which is my major interest.

So now I would like to plot in a single Graph all the 15 groups in a way where I can show the mean difference between the the 3 groups in each of the 5 conditions.

For instance, for the placebo condition I would be able to see:
Disease 1 - Placebo; Disease 2 - placebo and Disease 2 - Disease 1

And the same for the other 4 conditions.

Is it even possible?

Sorry if I wasn't clear enough making my point.





Adam Claridge-Chang

unread,
Aug 7, 2020, 3:16:42 AM8/7/20
to estimationstats
Hi Eduardo,

This would require ∆∆ plots, which we haven't yet implemented in DABEST or the web app (due to insufficient funding).

The first step would be to plot the effect sizes of each treatment against their controls, i.e. six 3-group Cumming plots, each with two effect sizes (disease 1 and disease 2), for a total of 12 effect sizes . This can be done with the current tools.

The second step would be to plot the ∆∆ of each drug versus their corresponding placebo control, e.g. ∆ for Drug A Disease 1 minus ∆ for Placebo Disease 1. These ∆∆ values would give you the size of the effect of each drug on the disease, accounting for placebo effect on the disease sufferers for a total of ten ∆∆ effect sizes. Comparison of the different drug effects could then be done by visual inspection of the ten ∆∆ curves. Not that this plot cannot be done with the current tools, and would require custom coding.

I hope I understood your question, sorry we can't provide the second part at this time. If you go through this Group, you'll see our responses to others who have requested ANOVA-type designs.

best,
Adam
Reply all
Reply to author
Forward
0 new messages