If you insist on doing this analysis yourself, you might want to get hold of the
book, Logistic Regression, by Kleinbaum.  Actually, you should consult with a
biostatistician if you want your results to be valid.

-Jay

Ronan O'Leary wrote:
> 
> Hi
> 
> I hope someone here may be able to point me in the right direction.
> 
> I am an undergraduate medical student doing a project which aims to look at
> whether information obtained in the diagnosis of certain forms of cancer can
> predict the presence of lymph node metastasis.
> 
> A "core biopsy" of tissue is removed from the cancer lump either using
> imaging techniques or palpation to ensure that the needle passes through the
> tumour. These pieces of tissue are examined by a pathologist who determines
> whether cancer is present and grades it accordingly. I would like to ask the
> following question with the information obtained: Is there any information
> in the core which can either predict the presence of disease positive lymph
> nodes or the absence? Lymph node disease is detected at subsequent surgery.
> I also hope to combine this with measurement of tumour size by clinical or
> radiological techniques.
> Therefore the fields I am using from my data set are:
> 
> Positive Lymph Nodes (yes/no or 1/0); grading criteria A (a numerical scale
> 1-3), B (1-3), C (1-3) and tumour size (diameter in mm).
> 
> I have been advised to use binary logistical regression and have been
> looking at Minitab, but have not really got that far despite reading every
> text book I can find with a chapter on this. I am beginning to think that
> the key problem is that each variable is not binary, they all have 3 values
> and the size can be a wide range, however, I can categorise this.
> 
> Basically I want to identify if any of the criteria alone or preferably in
> combination can predict lymph node disease. How do I interpret the results
> from Minitab. Is it possible to convert the results to % so for example I
> could say A1 B<2 C1 and tumour size less than 5mm means you are 5% likely to
> have mets.
> 
> Is there anyone who is able to help me out, it shouldn't take more than a
> couple of emails.
> 
> Thanks in advance
> 
> Ronan
> 
> P.S. Help on the best way to assemble the data and run the dialogues in
> Minitab would also be a great help.
> 
> ---
> 
> Checked by AVG anti-virus system (http://www.grisoft.com).
> Version: 6.0.344 / Virus Database: 191 - Release Date: 03/04/2002
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to