Re: Need to fit a trendline to a Negative Exponential set of data

  • Thread starter Thread starter oldman
  • Start date Start date
O

oldman

I have a number of data points that appear to fit a negative
exponential model. When I graph the data in Excel that is how it
looks. However, the trends that Excel allows do not appear to fit
negative functions. Can someone tell me how to modify the trend
fitting capability of Excel to handle this data.

Please post detailed steps. I am looking a cancer data that I need to
model and display.

Thanks,

oldman
 
Thanks for the response. It has been more than 20 years since I
worked with this stuff. What I am attempting to model the effect on
cancerous plasma cells of the drug Cytoxan. The patient has Multiple
Myeloma and has under gone two transplants and a stem cell boost. The
patient had relapsed again I is under going a new chemo regimen of
Dex, Velcade and Cytoxan. We so far have five sample points of 8.58,
5.80, 2.04, 1.42 and 1.28. The data looks like it is a decaying
exponential. The last two points have brought the patient into the
normal range but now appear to be tapering off. Each sample point is
the cancer marker for that week.

I can not seem to get the screen shot for function argument.

Though at this point there is not sufficient data to model this
situation I believe I am on the right track. I have never worked with
this capability it Excel so the reason for the post.


Hopefully, you can give me more help with how to use this capability
in Excel.

Thanks,
oldman
 
Though at this point there is not sufficient data to model this
situation I believe I am on the right track. I have never worked with
this capability it Excel so the reason for the post.
Click to expand...


List a set of numbers for an axis. List more. List them all! :-)


Let me play with the data, then you can label them however you like.

Also if you see it (one of my solutions)the way you want it, you would
see the chart characteristics as well,and be able to use it in the
future..
 
Hi thanks for the response. I will display what data I have which are
the first five points for the last five weeks of data since the chemo
protocol changed.

Week Kappa freelites
0 8.58
1 5.80
2 2.04
3 1.42
4 1.28
5 1.15
6 1.05
7 .99
8 .89
9 .81
10 .78


This should be enough of a start. The normal range if the disease is
under control is .33 to 1.94. These drugs are not a cure but only a
means to keep things under control. This is the sixth treatment plan
she has tried since being diagnosed almost five years ago.

Again we really appreciate the help.

oldman
 
Hi thanks for the response. I will display what data I have which are
the first five points for the last five weeks of data since the chemo
protocol changed.

Week Kappa freelites
0 8.58
1 5.80
2 2.04
3 1.42
4 1.28
5 1.15
6 1.05
7 .99
8 .89
9 .81
10 .78


This should be enough of a start. The normal range if the disease is
under control is .33 to 1.94. These drugs are not a cure but only a
means to keep things under control. This is the sixth treatment plan
she has tried since being diagnosed almost five years ago.

Again we really appreciate the help.

oldman
Click to expand...

Where are the negative exponents?
 
That exponent is one problem. I only have the time series that shows
the drug mix is working as depicted by the decline in Kappa Freelite
values. Is it possible to fit an equation to the data and thereby
determine the exponents. Its been over 20 years since I had to work
with this stuff. Until the wife got diagnosed I thought I was done
with this kind of stuff.

When she had other chemo rounds the curves looked like this data until
she would relapse. The reason why this data decays so slowly is that
the key drug is oral so the response is slower. As you can see from
the data this treatment plan is now really five weeks old and going.
When she had infusion chemo the decline in Kappa Freelite values would
occur over four or five day and she would be totally miserable for up
to three to four months.

Not sure if this helps you to help me.

thanks,
oldman
 
Herbert,

First thanks for your efforts and help.

Now for a few questions if you do not mind.
1-In chart 1 was that trendline computed by Excel 2007?
2-Is it possible to have solver compute residuals? I did this
by hard for the current data and results. If we took natural logs of
the observations would we then be able to compute a linear regression
model with errors and R^2 etc?
3-Where is the offset computed?
4-Going forward the data for days 35 will become real with
this weeks blood test. Can I substitute the real Kappa value for the
guestimated value and rework the model?
5-Working in the sheet I tried to copy and paste formulas but
using RxCy format only copied constants. What needs to be done to use
normal Excel operations?

Thanks again,

Paul
 
Herbert,

Thanks once more.

I see what you have done. Now I need to learn how to do this without
bothering you.

Too bad I can not send you the data that we have since Dec. 30, 2008.
What we are looking at is data that showed her in remission for a
while. Then the drug regimen she was on loosing its effectiveness
which resulted in a relapse and now a new treatment plan which is
getting the disease under control. After looking at the real data
for her disease we are hoping that the flat portion of the curve from
your model will truly extend. Should that be the case then we would
have a new situation where the readings vary randomly within the
normal range until this set of drugs loose their effectiveness.

Returning to your analyses for sheet 2 how did you compute Ao, t1/2,
k, and the b offset?

For the regression model the results are quite significant. I assume
you are fitting to the model from sheet2. Here again I need to figure
out what you setup and how you did it.

All help is really appreciated.

Paul
 
Herbert,

Thanks once more.

A few more questions:

1- What happens when I get actual data beyond day 70? Every Tuesday
there is a blood test to measure Free Kappa Lite chains to see if the
treatment protocol is still working. This Friday we will get the real
value of Kappa for day 35.

2- What is the effect of additional data on the coefficients Ao, t1/2,
k and the offset b. I changed the value for day 35 to see what would
change and so far it only changed the value r4c7.

Paul
 
Back
Top