T-test for comparing test and control means for several experiments

[Trey1]

Hi,

I am trying to test the difference in means for several experiments in the same Alteryx flow. Each experiment has a test and control group indicated by a dummy variable. If Population and Value are summed to the experiment and testing group level, the data look something like:

Experiment	Group	Population	Mean_Value
1	Test	15	100
1	Control	17	199
2	Test	75	164
2	Control	36	356
3	Test	52	853
3	Control	46	235

The 'Test of Means" tool will give the t-score, etc for a the test vs control, but it doesn't have any way to group on experiment. I need to know the p-value for if the test value is significantly different than the control value for each experiment. 

The only way I have figured out to do this would be to separate the data be experiment, use the Test Means tool to give the p-value, and rejoin the data. That is really cumbersome and not dynamic as I add new experiments to the bottom of the data.  

Any help would be greatly appreciated! If only the Test of Mean tool had a "groupby" function. 

Thanks,

Trey

[T_Willins · Accepted solution]

Hi @Trey1,

A batch macro will give you the group by you are looking for and will allow for additional experiments to be added to the data.

[Trey1]

Thank you for the super fast response with a macro!

I'm updating my Alteryx to use it, but it will take some time.

Will this macro automatically do the test for every experiment or do I have to create a new filter for each one? 

[T_Willins]

Hi @Trey1,

This batch macro will filter each data set to one experiment number at a time and process that data through the Test of Means, then repeat for the next experiment number and union the output of each separate experiment number's result.  I added a Formula tool to label each row in the Macro output.  As long as each experiment is labeled with a unique number, you can keep adding data and this macro should run.