Discover MakerZone

MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn more

Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply Today

Thread Subject:
Need Help with multi-dimensional results analysis

Subject: Need Help with multi-dimensional results analysis

From: Eric Diaz

Date: 19 Aug, 2011 02:45:11

Message: 1 of 9

Dear Matlab Community,

I am having great difficulty trying to summarize / analyze my results. The problem is making my brain hurt a bit actually because I can't seem to wrap my head around how I want to start or finish.

I conducted an experiment for model validation by controlling data creation variables, data acquisition variables, and data fitting models. All in all, there are about 4 variables and 3 models. I want to see how each of these variables effects the accuracy and precision of the models. For example, # samples vs. true parameter value, or sample spacing vs. true parameter value, or SNR vs. true parameter value. I repeated these experiments over 100 times for each one, for statistics.

Now, my parameters matrix is (7,9,9,11,3,5), with the last 5 being the 5 parameters being calculated, and the first four dimensions being 4 variables for the experiment, and the dimension with 3 being the 3 different models.

I know the true values of the parameters, since I created the data, and so want to summarize the result for each parameter against the true value, for each of the experimental changes.

Can anyone help me wrap my head around this problem and suggest some approaches for summarizing the results, please?

Thanks,

Eric

Subject: Need Help with multi-dimensional results analysis

From: Eric Diaz

Date: 19 Aug, 2011 21:58:09

Message: 2 of 9

Can anyone suggest a tag / keyword that may help this post be seen by people that may be interested in this problem?

Thanks,

Eric

Subject: Need Help with multi-dimensional results analysis

From: ImageAnalyst

Date: 19 Aug, 2011 22:22:54

Message: 3 of 9

Sounds exactly like the "design of experiments" analysis our
statisticians do in JMP. I'd suggest you contact your organization's
statistician for help. They have all kinds of fancy ways they can
analyze this in JMP.

Subject: Need Help with multi-dimensional results analysis

From: Eric Diaz

Date: 21 Aug, 2011 01:29:08

Message: 4 of 9

Thanks for the suggestion ImageAnalyst. Unfortunately, I don't have paid access to a statistician. The only statistician I have access to is not part of my group and she is often quite busy. I normally don't ask many questions of her since it is above and beyond her duty to help.

Is there any other suggestions that someone can make?

Subject: Need Help with multi-dimensional results analysis

From: Greg Heath

Date: 21 Aug, 2011 03:28:48

Message: 5 of 9

On Aug 20, 9:29 pm, "Eric Diaz" <eric.d...@gmail.com> wrote:
> Thanks for the suggestion ImageAnalyst.  Unfortunately, I don't have paid access to a statistician.  The only statistician I have access to is not part of my group and she is often quite busy.  I normally don't ask many questions of her since it is above and beyond her duty to help.
>
> Is there any other suggestions that someone can make?

You can't ask her if your company has JMP avalable for you to use?

Greg

Subject: Need Help with multi-dimensional results analysis

From: Eric Diaz

Date: 21 Aug, 2011 19:30:12

Message: 6 of 9

Oh, sure, I can ask that. But I assumed JMP was just similar to other statistics software and I think she generally works with R. And I also assumed that if we did have JMP i would probably need to ask her for help. I know someone who works for Stata or SAS or one of those companies. Do you think they might be able to help?

Subject: Need Help with multi-dimensional results analysis

From: Eric Diaz

Date: 25 Aug, 2011 01:12:13

Message: 7 of 9

Hi,

I just wanted to let you guys know that I asked about JMP and she wasn't familiar with it. She said all she uses is R.

Is there any other suggestions as to who I might turn to, what keywords I might use to attract interest, or what other options I might have.

Thanks.

Subject: Need Help with multi-dimensional results analysis

From: John Barber

Date: 25 Aug, 2011 02:16:10

Message: 8 of 9

"Eric Diaz" wrote in message <j347hd$75n$1@newscl01ah.mathworks.com>...
> Hi,
>
> I just wanted to let you guys know that I asked about JMP and she wasn't familiar with it. She said all she uses is R.
>
> Is there any other suggestions as to who I might turn to, what keywords I might use to attract interest, or what other options I might have.
>
> Thanks.

I believe that you can download a free trial of the JMP software, but I don't know how limited it is. A competing product that I have used more recently than JMP is StatGraphics - they also might offer a free trial version. What both of these packages offer compared to MATLAB is a set of good GUI-based tools for DOEs and subsequent analysis. They make it very easy to make a good set of plots to analyze your results (correlations, residuals, probability plots, etc.). However, you still need to know enough to interpret the results.

If you know your statistics reasonably well, you can just do it all in MATLAB, if you have the Statistics toolbox and can live with less GUI interaction. You also need to know what kinds of plots are appropriate for your model/data, as well as what statistical tests you might need to run.

If you aren't familiar enough with statistical analysis to already have an idea of the kinds of things you need to do to design a proper experiment and interpret the results with reasonable confidence, I recommend obtaining a good textbook on the subject, as well as a sit-down consultation with an expert.

-John

Subject: Need Help with multi-dimensional results analysis

From: Eric Diaz

Date: 25 Aug, 2011 20:57:10

Message: 9 of 9

Thanks John. I've been looking at the videos tutorials for DOE, available on the stat graphics website, http://www.statlets.com/videos.html , and they are fantastic! I think I may be able to devise a way to do this in MATLAB or I may just try the 30day free trial of stat graphics. Thanks a bunch.

> I believe that you can download a free trial of the JMP software, but I don't know how limited it is. A competing product that I have used more recently than JMP is StatGraphics - they also might offer a free trial version. What both of these packages offer compared to MATLAB is a set of good GUI-based tools for DOEs and subsequent analysis. They make it very easy to make a good set of plots to analyze your results (correlations, residuals, probability plots, etc.). However, you still need to know enough to interpret the results.
>
> If you know your statistics reasonably well, you can just do it all in MATLAB, if you have the Statistics toolbox and can live with less GUI interaction. You also need to know what kinds of plots are appropriate for your model/data, as well as what statistical tests you might need to run.
>
> If you aren't familiar enough with statistical analysis to already have an idea of the kinds of things you need to do to design a proper experiment and interpret the results with reasonable confidence, I recommend obtaining a good textbook on the subject, as well as a sit-down consultation with an expert.
>
> -John

Tags for this Thread

What are tags?

A tag is like a keyword or category label associated with each thread. Tags make it easier for you to find threads of interest.

Anyone can tag a thread. Tags are public and visible to everyone.

Contact us