Allowed calorie intake...?

Turn your data into smooth curves automatically with TableCurve 2D by
Systat Software, Inc.

http://www.systat.com/products/TableCurve2D/

http://groups.yahoo.com/group/TableCurve/









"GaryG" wrote in message ...
"Heywood Mogroot" wrote in message
om...
"GaryG" wrote in message
...
"Bob" wrote in message
While researching this, I realized that my digital scale could provide
some
answers. So, I started weighing myself every day. But, after doing
this
for a while, I noticed a problem...weight changes quite a bit from day
to
day. Variations of 2-4 pounds are not uncommon (due to many natural
causes,
in addition to that container of Haagen-Dasz!).

Fortunately, the problem of "variability" in data has been addressed in
other fields.

And diet books!

I found this free diet book:

http://www.fourmilab.ch/hackdiet/www/hackdiet.html

(the infamous "Hacker's Diet")

to be an excellent educational source (and the basis for my diet
philosophy).

Scientists and stock market analysts use a technique called
"linear regression" to filter out the daily ups and downs in their data.
So, I decided to write a program to do this for weight management.

Funny, I've been thinking of marketing a diet program too, and
checking over your site for the first time I see a lot of common
ideas! (I like your calendar layout idea -- very smart).

It's a weight and health diary called "WeightWare"
(http://www.WeightWare.com).

I also must say the overall quality of design of your work is very
apparent. Bravo! (and this is coming from a full-time programmer / UI
designer!).


BTW - if you do decide to download and try WeightWare, I would welcome your
feedback on it. You can contact me via email at garyg -at-
shastasoftware -dot- com.



I hesitate to torpedo your work here, since it is so good, so I'll
refer anyone who is interested in this important aspect of dieting
(maintaining statistics daily) to my other post in this thread that
I'll write now.

...

You can download a free, fully functional 45-day trial version from the
website to see if it meets your needs. Sorry for the somewhat lengthy
plug,
but I've been working really hard on this program, and just officially
released it this morning (yeah!).

Hah, like I said, clearly an excellent job.

Just a quick note...I've released version 1.3 of WeightWare. Along with a
couple of bug fixes for metric units of measure, it includes two new graphs
that you might find useful.

The "Weight and Moving Average" graph shows your daily weight, along with an
exponentially smoothed moving average (using a .75 smoothing factor), for
any time frame that you choose.

The "30 Day Calorie Trend" graph shows each day's calorie deficit/surplus
averaged over the preceding 30 days (using linear regression). This 2nd
order graph allows you to quickly determine if your rate of weight loss is
increasing or decreasing, allowing you to make changes in eating/exercise
patterns before getting off track. As far as I know, no other program
offers this feature.

GG
http://www.WeightWare.com
Your Weight and Health Diary

"Heywood Mogroot" wrote in message
om...
"GaryG" wrote in message
...
+ if you plot the expected rate loss, you will have to support
variable loss rates, since it is common to switch gears on a diet
(partway in I went from 875 kcal/day deficit to 1000 kcal/day, but
perhaps later I might want to start backing off a bit in a month or
two).

+Your trend plotting is very linear; I think too much information is
being lost with your function. I prefer John Walker's formula (with a
0.75 smoothing factor), since it is more usable on a day-to-day basis.

It's linear by design. The weight trend shown (for whatever time period
you
select) is based on "least squares linear regression". The same
technique
is mentioned on page 294 of the Hacker's Diet.

All the "hacker's diet" followers on the web use the moving average
method of graphing, AFAIK.

During development, I experimented with also displaying an exponentially
smoothed moving average, but it didn't seem to add much value.

The least squares just seems like it is drawing a single line to fit
the data. As I tried to imply, this is fine for looking back at a
period of time, but doesn't seem to provide any useful information
going forward, since important recent trends are apparently getting
chopped off with the single best fit curve fitting.

The eye can determine the linear trend easily enough, but when I look
at my chart, it's the bumps and dips in the moving average that tell
me how things were going at that period of the diet.

By fitting all the datapoints to a single line, you've basically
reduced all information for that period down to a single datum: the
average rate of loss over the entire period.

Now that I
think about it, I think it depends on the time frame you're looking at.
For
shorter time frames (say, 1-3 months) I think the linear "weight trend"
provides more useful information.

Disagree. See above. The most important thing is what's happening this
week, and what the weight will likely be tomorrow if present trends
continue.

For longer time frames, I think the
moving average would give a better view of changes in weight. I'll
probably
add a "Moving Average" view to the graphs in the next version.

recommended

I'm also thinking of adding a "30-day Weight Trend" view...this would
show
the 30-day linear regression value for each date (i.e., the 30-day least
squares daily caloric deficit). Changes in this graph would allow you
to
quickly see if your rate of weight loss is increasing or decreasing.

yeah 2nd order trends are useful. In my spreadsheet I color-code the
daily gain/loss red if it is less than planned (0.286 lbs/day in my
case). My other (sooper sekrit proprietary LOL) numbers are also
2nd-order related, and similarly color-coded. Too many reds in a row
and I know something's not working.

+ minor UI thing on the calendar, the weekly summary cells should not
be so similar to the out-of-month cells. I'd also color-code 'up' days
in red, to better signal them.

I'm hesitant to color-code like that, for several reasons. Firstly, the
monthly min/max weights are already color coded (green and red,
respectively).

yeah I saw that and first assumed you we're color-coding like me :0

I also don't want to overemphasize the "up" days like that
because they occur naturally, even during a successful weight loss
program...coloring them red could add unnecessary stress to the
weigh-ins
(the status panel at the bottom of the month shows how many days were Up
vs.
Down, however).

Here's where I think you're missing the crux of the Hacker's Diet...
the exponential smoothing function filters this up & down business
already, so if & when you're in an "up" situation it means the
trending is pretty out of control. Of the 75 days (rows) in my diet
worksheet, the daily gain/loss column's got 7 "red" figures (days
where I the daily scale weight exceeded the previous calculated
average) located in 3 groups.

In fact, none of my calculations (other than the averaging function)
even refer to the scale weights, everything is using the weighted
averages. And when I post here I always refer to the weighted average
as my current weight. The purpose of the weighted average is to buffer
noise, and at least for my weight loss pattern seems to be working
well.


Heywood
232/208/182