Need a little help with a very simple sensitivity analysis

Hello everyone,

I have what I think should be a simple questions.

I have some experimental data which is dependant upon two variables and I would like to know which of the two is more important on the output. The example I am thinking of to keep this general would be: I do 2 different exercises a day for 100 days and I record how much I drink each day. So, A = number of pull-ups done each day (between 1 and 10), B = number of pushups done each day (between 1 and 10), and Y = water drunk each day.

What I would like to asses is which of the two exercises has a great effect on the water drank each day.

Can anyone point me in the direction of some very simple analysis techniques which would allow me to quantify which exercise (A or B) had the bigger influence on Y?

Apologies for asking such a simple question but I found myself deep into latin hypercubes and 50+ parameter spaces which all feel like over-kill.

I'm not asking for anyone to do this for me, but I just am not sure which type of analysis is needed. Is ANOVA suitable? I've seen that Sobel is possibly suitable, and that perhaps the Sensitivity Analysis toolbox is worth looking into?

Any help would be very appreciated!

Rob

EDIT: How about cutting me some slack John? I have tried to make it general. And I'm clearly not asking someone to do it for me. I'm asking for some help, isn't that was this is for!?

[José-Luis]

ANOVA?

Although I must admit I don't really follow your example.

[John D'Errico]

Answers is here to answer questions about MATLAB. That is why it is called MATLAB Answers. Is there any aspect of your question that asks a question about MATLAB? No. I've left it open now only because someone chose to provide an answer.

[Jan]

@Robert: You can clarify the connection to Matlab during the discussion. E.g. if you found a method you prefer, talk about the application in Matlab.

Your example looks very artificial. You can be sure not to find any significant correlation. But you can run a student t test and decide for the "better" p value. Many studies are based on tortured statistics.

[Robert Malkin]

Hi Jan, thank you for the input.

I am trying to perform this analysis using the fitlm function in Matlab. I know that there is likely no correlation in the example I have given, but I was trying to keep it general. In reality I have 600 experimental data points which are based on 3 variables. In essence, 3 dials on a black-box and I'm trying to understand which of the 3 dials has the greatest effect on the output. I hope that makes sense!

Answer 1

[the cyclist]

If you have the Statistics and Machine Learning Toolbox, I think you should be able to use fitlm to do this.

[Robert Malkin]

I do indeed have it. Thank you. I will read up and report back.