Happy Homemaker, Ph.D.

A novice homemaker's attempts to use her engineering Ph.D. to serve her family

The Oatmeal Scotchie Experiment – The Technical Side

on March 21, 2013

Disclaimer:  A couple of close friends have told me that sometimes my posts are such that they don’t understand them.  One friend who said this said she understood that I was “venting.”  I thought this was a great description, not venting as in complaining, but in needing an outlet to speak “my language” of applied stats, optimization, operations management, and process improvement.   I’m still trying to determine what audience I most want to reach with my stories, so for now, I write a little bit of everything.  So, if you’re one that I lose when I pull out my fun-for-me technical vocab, this post may not be for you.  Don’t worry.  Just skip it and tune in next time.  🙂

Designing an experiment properly is not an easy task.  I didn’t do a proper job of it when I explored my questions about the cookies this week, but I wanted to discuss some of the considerations a “real” experiment of this type would need to address.

First, what is the response variable?  In this case, I was looking at cookie quality, but how could that be measured?  I could have had multiple testers (Les, thanks for volunteering!) who would rate the cookies on a scale.  (Not my top choice as a non-continuous output variable, though.)  Since I was interested in voids, I could have taken samples and counted holes per square inch or per cookie somehow (if “holes” was well-defined to be consistently judged).  Since I have some idea of what the cookie should look like, I may have measured height (to gauge a flat or mounded cookie with calipers or something).  See, it’s getting a bit tricky already (or at least time-consuming), isn’t it?

Next, what factors did I want to explore?  Since the different amounts of flour and oatmeal were part of the recipe, I first wanted to look at a mixture experiment, but that wasn’t a good fit for this.  (Ooo!  Perhaps I can do a mixture experiment to optimize homemade play dough or something later…)  So, using the flour/oatmeal combinations given and regarding them as simply 2 choices seems best.  I didn’t think there was much to be gained by trying the in-between combinations.  Next, butter vs. shortening was of interest.  Also, I wondered about the pans, thinking there may be a butter-pan interaction.   Perhaps the refrigeration of the dough before baking would also have an effect.

With these four factors, I would have to make a minimum of four batches of dough – and that design would include some aliasing or confounding (Which means the statistics would a factor would be significant, but you can’t be sure if it’s a main effect or a two-way interaction, for example, in what’s called a Resolution III design.).  That number of experiments is not bad, really, but it’s more than I have room to store at one time.  And frankly, that’s a little more hassle than I’m up for right now with the little ones.

Minitab gives me this half-fraction factorial design (Resolution IV):

StdOrder RunOrder Blocks Recipe Butter Pan Cooled Dough

4

1

1

1

1

1

-1

1

2

1

1

-1

-1

-1

6

3

1

1

1

-1

1

2

4

1

1

1

-1

-1

8

5

1

1

1

1

1

3

6

1

1

-1

1

-1

5

7

1

1

-1

-1

1

7

8

1

1

-1

1

1

With an alias structure of:

I + ABCD
A + BCD
B + ACD
C + ABD
D + ABC
AB + CD
AC + BD
AD + BC
 

I’d also want to be considering blocking in the experiment, because I don’t see how it can be avoided without increasing the “costs.”  Blocking would simply mean, for example, that I’d be baking more than one sheet of cookies at the same time.  The cookies baked in that particular oven cycle have that same treatment that may differ in some ways from the other baking batches.  Identifying something like this helps assign known possible variation to this factor, rather than just adding it to the random variation (e.g. everything else not accounted for with the chosen factors).

There may also be a bit of an ethical dilemma that educational scientists and social scientists are really more familiar with than those of us in engineering or the “hard” sciences.  If you know a particular combination will short-change the subjects (in this case, make the cookies bad), should you go through with that “treatment?”  Okay, so burning cookies or making them crumbly to prove an experimental effect really isn’t much of an ethical issue, but it’s a good question to consider.  Even in the hard sciences, can your experiment be set up to avoid known failures?  (It can.  It involves changing the design space…)

So why, if I love experimental design – and cookies – did I not take this more extensive approach?  (This experiment would yield about 250 cookies.)  Perhaps some of the first questions an experimenter needs to ask is what is he or she interested in learning, what is the available budget (time, money, resources) for the study, and what is the value of the learned information.  Every world – the business, the lab, the home – must have priorities for the many demands facing it.  These limitations are part of what has driven the development of computer experiments that look to optimize the efficiency of designs, meaning finding designs that give you the most “bang for your buck,” or the most information for the minimum number of “runs.”

My thoughts were that I wanted to gather some information with some quick trials while avoiding a one-factor-at-a-time (OFAT) approach (which would likely make me fat, too), so I was interested in a sort of screening experiment.  For minimal costs, (the time it took to make one batch of cookies), I found a way to make the cookies stay together, discovered butter wasn’t the only issue, and found my insulated pans weren’t causing big problems, either.

The results? Delicious cookies, a happy family, and a more confident mama.  My husband and girls wouldn’t be super-impressed with my main-effects and interactions plots from Minitab or the p-value for any statistically significant effects.    Really, my oldest is still 2.  Maybe next year I’ll start showing her that stuff.  😉

Advertisements

2 responses to “The Oatmeal Scotchie Experiment – The Technical Side

  1. indytony says:

    Great news. Perhaps if Betty Crocker had engaged in such an analysis the world would be a better place.

  2. Amy Roberts says:

    Dana – I read your post out loud to my husband because they make me laugh so hard. I can’t even begin to tell you how much I enjoy your outlook on homemaking. LOVE IT! Keep writing this stuff…there is an audience. 😉

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: