ietoolkit
ietoolkit copied to clipboard
worldbank
iebaltab : Order of the difference
It would be great if we could have the mean difference as being (2) - (1) / (3) - (1) and (3) - (2). As (1) is usually the control group, I think it makes conceptually more sense to have treatment - control than the opposite to have a positive result when treatment is effective.
That's a very good point! I completely agree, and do not know why it isn't like that already. Thanks for bringing this up!
If this command would not already have been published I would definitely do as you suggest. But since that is not the case I want to think about it and discuss this with @luizaandrade and @bbdaniels . The problem is if someone already is using this for a table, and make a reference to a number, ex. -1.8, in the text, updates the command, and now that number is 1.8.
A compromise would be to add what you are suggesting as an option, but I think your suggestion should be the default, and the option to switch this feels like a patchwork kind of solution that we might end up regretting later. :confused:
I feel like we're getting to the point where major revisions and backwards compatibility is going to start being a bigger issue in general. Opened #179 to discuss in detail.
This is solved in version v7.0 by introducing the command control(). That indicates one group as control and test all other groups against it, and then flips the sign so positive difference means that treatment has a higher value than control.