I agree with you about the dangers of normalizing diabetes data, however, I
was courious about how to do the process and wanted to see what the results
would be. How would you do the normalization if the data were not linear?
Years aga I ded research with beef calves where we measured their weight all
on the same day and thne normalized their weight to 1 year old. Ther weights
were nor linear and I can't remember what statistical model was used to do
the normalization.

"FinRazel" wrote:

Sorry, I think I accidentally reverse sorted one column independently of the
other, finding a direct, rather than an inverse correlation. For the inverse
equation:
X = (Elapsed Time)*(Meaured Glucose)/(Time=2 hours)

But, I think you should follow what joe says, and measure your glucose
levels at exactly the same time. You don't want to be estimating numbers
based on an assumption of linearity that is probably not correct.

For the sake of experiment, though, you could test, for example, at 1.6,
1.8, 2, and 2.2 hours in a given day and see how well the transformed values
match your 2 hour time point.
--
Anne Murray


"FinRazel" wrote:

The data you have given follows a linear relationship... if you plot Glucose
level vs. time, on a scatter plot or line graph, you can see this. Since
this is the case, the linear correlation applies: X1/Y1 = X2/Y2. To
normalize, the time you want (X2) is always equal to 2. Y1 is your recorded
Glucose level, and X1 is your recorded time. Your normalized Glucose level
is Y2. In short, if you have time in column A and glucose level in B, fill
this formula into C (normalized): =2*B2/A2

Of course, you should get more data points to see how accurate this is.

Cheers!
--
Anne Murray


"k1ngr" wrote:

I am diabetic and test my blood glucose level 2 hours (approximately) after
eating.
Since I don't always test exactly after 2 hours, how can I normalize the
glucose reading to 2 hours? Below is a sample of my data:

ELAPSED TIME GLUCOSE Normalized
1.9 150 ?
2.2 110 ?
1.7 160 ?
2.3 100 ?
2.0 130 ?

Thanks,
k1ngr

I looked up blood glucose level on wikipedia, and the article has an
idealized curve showing blood glucose with time. Blood glucose levels
generally increase throughout the day, with major and minor spikes after
meals. This is, of course, nonlinear, but there are intervals that could,
perhaps, be treated as linear. On the other hand, with enough data points,
you might be able to obtain the rate constants for the polynomial function
(2nd order) representing each spike, which might lead to a more reliable
transform.
--
Anne Murray


"k1ngr" wrote:

I agree with you about the dangers of normalizing diabetes data, however, I
was courious about how to do the process and wanted to see what the results
would be. How would you do the normalization if the data were not linear?
Years aga I ded research with beef calves where we measured their weight all
on the same day and thne normalized their weight to 1 year old. Ther weights
were nor linear and I can't remember what statistical model was used to do
the normalization.

"FinRazel" wrote:

Sorry, I think I accidentally reverse sorted one column independently of the
other, finding a direct, rather than an inverse correlation. For the inverse
equation:
X = (Elapsed Time)*(Meaured Glucose)/(Time=2 hours)

But, I think you should follow what joe says, and measure your glucose
levels at exactly the same time. You don't want to be estimating numbers
based on an assumption of linearity that is probably not correct.

For the sake of experiment, though, you could test, for example, at 1.6,
1.8, 2, and 2.2 hours in a given day and see how well the transformed values
match your 2 hour time point.
--
Anne Murray


"FinRazel" wrote:

The data you have given follows a linear relationship... if you plot Glucose
level vs. time, on a scatter plot or line graph, you can see this. Since
this is the case, the linear correlation applies: X1/Y1 = X2/Y2. To
normalize, the time you want (X2) is always equal to 2. Y1 is your recorded
Glucose level, and X1 is your recorded time. Your normalized Glucose level
is Y2. In short, if you have time in column A and glucose level in B, fill
this formula into C (normalized): =2*B2/A2

Of course, you should get more data points to see how accurate this is.

Cheers!
--
Anne Murray


"k1ngr" wrote:

I am diabetic and test my blood glucose level 2 hours (approximately) after
eating.
Since I don't always test exactly after 2 hours, how can I normalize the
glucose reading to 2 hours? Below is a sample of my data:

ELAPSED TIME GLUCOSE Normalized
1.9 150 ?
2.2 110 ?
1.7 160 ?
2.3 100 ?
2.0 130 ?

Thanks,
k1ngr