Effect size in gctest

Dear all,

I'm using gctest to test if there is any causation between three variables. The table is a 300 x 3, with 3 columns for each of the variables (let's call them A, B and C). Let's imagine that I find a significant granger-causality between A and B (A predicts B) and between A and C (A predicts C), how can I know the effect size of these two relationships? The test statistic is not useful here, among other reasons because I also need the sign of the causality (e.g., A predicts B, but this could be due to either A promotes or inhibits B).

Thanks a lot for your help!

Answer 1

[Avadhoot]

Hi Laurent,

I understand that you have already performed the Granger causality test between variables A and B, as well as between variables A and C. To calculate the effect size for this causality, you can use the "meanEffectSize" function. This function calculates the mean-difference effect size between two samples. Please refer to the below documentation to know more about the “meanEffectSize” function:

https://www.mathworks.com/help/stats/meaneffectsize.html?s_tid=doc_ta#d126e736622:~:text=0.79403%20%20%20%200.58717%20%20%20%20%201.0009-,Compute,-Median%2DDifference%20Effect

Here is a reference code for the same:

Effect = meanEffectSize(A, B); 

Here's what this line does:

1. Variables A and B should be numeric vectors representing the first and second samples, respectively.
2. The "Effect" variable will contain a table with a row for each calculated effect size, and columns for the effect size value and confidence intervals.

For more information about Granger causality test, refer to the documentation below:

• gctest function – Contains information about the Granger causality test function.

I hope it helps,

Regards,

Avadhoot.

[Laurent]

Thanks Avadhoot for taking the time to answer my question.

I'm not sure however what should be the input to meanEffectSize. Especially since I'm looking at effect sizes of specific relationships (A predicts B and A predicts C for example). Would you know which output of gctest I should input?

Thanks!

[Avadhoot]

Hi Laurent,

You do not need to pass the output of gctest to the "meanEffectSize" function. Instead you need to pass the actual variables from your table. For example, your table contains 3 variables A, B and C. So to find the effect size of the A predicts B relationship, then you must write the code as follows:

Effect = meanEffectSize(A, B);

The gctest function is just to establish the causality. To calculate effect size use the original variables A and B.

Hope this helps.

[Laurent]

Thanks for you time Avadhoot. However, if I compute meanEffectSize(A, B), I will get, as an effect size, the mean difference between A and B and not the effect size of A predicts B. To my understanding, meanEffectSize does not provide the effect size of causality, but I'd be happy to be provden wrong...

[Scott Stanslaski]

Laurent wondering if you ever found an answer to this question? I'm trying to do the same thing.