Re: Carbohydrates--plots and linear programming results.

[Guest guest]

> >>If you're looking for ways to evaluate or rank carbohydrates how about

> a cost benefit for nutrition vs. energy density or energy per serving.

>

> One of the ways I have been wanting to do, but havent gotten around to

> yet, was to plot fruits, veggies and unrefined complex carbs on one line

> by calorie density and another by fiber/carb ratio and a third by

> nutrient density (it would have to be 3D). Then split the graph into 4

> quadrants. IT would than be easy to rank the foods

>

> Anybody got the time, i got most all the numbers

It doesn't tell you much. Nutrient density and calorie density are

largely inversely proportional, as you might expect. The relationship

between nutrient density and proprotion of fiber as calories is not so

well-defined in whole foods, though. There is a trend for low or

zero-fiber foods to have low nutrient densities, but at higher

nutrient densities, one sees both relatively low- and relatively

high-fiber foods.

I have made some plots and run some linear programming models. There

are some mild problems with some of the data (I used SR13--some values

are missing but they still haven't been added as of SR 16). There were

also some problems with the plotting method. I used gnuplot, which

likes to see a full set of (x,y) values rather than the discrete

points that are derived from the nutrient database--this precluded

generation of a full surface, so I had to plot using points instead of

lines or contours.

You can see the data I used and the plots here:

http://www.atlantc.net/~ded/nutrient_density.html

I also wrote a vegan linear program to find the optimal solution to

the diet problem. It does not address B12, D, or long-chain omega3s,

which must be added separately.

For those of you who are not familiar with linear programming, this is

just a way to find the minimum or maximum of some objective (here,

minimize calories) while satisfying a number of specified constraints

(meet all DRIs and other specifications, which could be based on color

for phytochemicals, limits on antinutrients, etc.). The sky is the limit.

If you are not vegan, you can still get some useful information from

these results, and just add your 60 g or so of sardines in tomato

sauce or cow brains or whatever to address B12, D, and long-chain

omega3s. But, I do plan creating a model with animal source foods in

the future.

I used the free lp_solve code (in C) to generate results. There is a

group discussing the use of this model. There are many versions

available for a diverse set of platforms, including vb, .net, and

Excel and OpenOffice plug-ins (which I have not yet gotten to work).

The model is here (in plain text):

http://www.atlantic.net/~dec/vegan2.lp

and some results are here:

http://www.atlantic.net/~dec/opt.txt

There are some problems with the data I used: SR13 from

's website, which does not provide linoleic acid values. I

calculated them from the total fat, saturated fat, MUFA, and ALA

values, and they do not always match precisely with SR16 values.

However, I would like to point out that there remain a number of

problems with the SR16 data. Take, for example, almonds: the

saturated, MUFA, and pufa gram totals do not sum correctly. This is

only one example of a problem. There are many others.

As you might expect, when you run such a model, if you do not include

realistic limits for certain types of foods, it will give you the best

exactly according to what you specify, which might be something

ridiculous like 17 cups of watercress in the solution, which is just

about exactly what the first case comes up with. Even if you really

wanted to eat 17 cups of watercress, which I do not advise, you could

not afford it. It would cost about $20 for 1000 kcal of watercress in

my area. Cost is a big problem. If you consider the cost per calorie

of the most nutrient-dense food, rather than the cost per pound, you

can see that it can be a very expensive proposition to eat well in

this society.

Another issue that comes up repeatedly is the problem with redundant

(or largely redundant) food items. If you have several entries for a

particular type of food (raw, cooked, canned), the model will like to

choose all of them, if it likes that particular food, so it helps to

eliminate as much redundancy as possible ahead of time.

Future additions will include animal source foods, phytochemicals,

costs, genetic diversity, and antinutrient limits.

Carol

[Guest guest]

> sauce or cow brains or whatever to address B12, D, and long-chain

LOL! I just had to throw this in:

" Vital Glands provides peptones from glandular meats which have been

processed into a free flowing powder using a special 'low-

temperature' process to assure the maximum freshness and nutritional

biological activity for this food.

Made from the following glands:

Adrenal --- Lung --- Spinal Cord

Bone Marrow --- Lymphatic --- Spleen

Brain --- Mammary --- Stomach

Duodenal --- Orchic (bull testicles)

Suprarenal --- Eye --- Ovaries --- Thymus

Heart --- Pancreas --- Trachea --- Hypothalamus

Parathyroid --- Umbilical --- Kidney

Pituitary --- Uterus --- Liver --- Pineal

and others "

http://www.healingedge.net/store/more_dew_vitalglands.html

Logan