Back to the Top
Hello All
Is there any reference paper which discusses the various methods for
transformation of data and its implication in NONMEM analysis? Could
the group share their experience on data transformations? Especially
what aspects of nontransformed data should be looked into before
selecting the suitable transformation of data. In one of the analysis
I was observing that log transformation of data helps in getting
better estimates and the analysis is more stable. The data is after
infusion studies in patients who are also on several concomitant
medications. However, I was able to fit the data of healthy subjects
data with out any transformation.
What could be the possible reasons for these type of observations?
Thanks in advance for your time
Venkatesh Atul Bhattaram
Post-doctoral Fellow
University of Florida
Gainesville
Back to the Top
Venkatesh,
Our approach, when we also have intensive data, is to fit each individual in
the development of the appropriate structural model, and to look at
parameter distributions. Those that are right skewed get log transformed
for the population analysis. Parameters like Ka's, Tlags, & volumes usually
do not follow a gaussian distribution, and should be transformed. The thing
to remember is that a log transform on your parameters often helps, and
never hurts. With intensive data and something like maximum likelihood
estimation, you can get away without doing a log-transform...it's the
population approach, especially if using bayesian estimation, that requires
an accurate assesment of the parameter likelihood distributions.
I'll be interested in hearing the thoughts of the group...
regards,
patrick smith
buffalo, ny
Back to the Top
The log transformation will always help to stabilize the variance in
an analysis, even when the data are normally distributed. It also
helps with interpretation on a ratio scale as opposed to a
difference.
A general paper on the log transformation can be found in:
Keene ON (1995) The Log Transformation is Special Statistics in Medicine.
Alun Bedding
Senior Statistician
Clinical Pharmacology Scientific and Regulatory Expert
Statistics and Information Sciences
Eli Lilly and Company
Back to the Top
I consulted my colleague on this issue, here is his opinion on
transformation of data. Is it not a fact that we have to look at the raw
data and use transformation and weighting only under special condition
rather than a routine practice?
Regards,
Prasad Tata
-----Original Message-----
From: Allen Jr., John C
Sent: Tuesday, April 23, 2002 10:14 PM
To: Tata, Prasad N
Subject: Re: When to do transformation of data?
Dear Dr. Tata:
In my experience (30+ years), if the data are normal the log transformation
won't help the analysis much--if at all--and it won't hurt it, either. If
the data are skewed to the right, a log transformation will pull the upper
tail in and make the distribution less skewed, which will help the analysis
by increasing the power (sensivity for detecting a difference). It is true
that the log transformation can be used to allow interpretation of the
results on a ratio scale as opposed to a difference scale.
John C. Allen, Jr.
Manager, Biostatistics
Tyco/Healthcare Mallinckrodt
Back to the Top
Dear Venkatesh,
The Log transformation of emperical data is one of
the commonly used transformtaions in data analysis. An
useful book in that field is "Exploratory Data
Analysis" of John W. Tukey, Addison-Wesley Publ Comp,
1977.
According to my experience, data
transformations are useful when using parametric
methods for analysing of data with non-Gaussian
distribution. When using IT2B and/or NPEM algorithms
(implemented in USC*PACK developed by the group of Dr.
Roger W. Jelliffe) you can perform your data analysis
without doing data transformations. In this context I
agree with the statements of Patrick Smith.
Yours sincerely,
Dimiter T.
Dimiter Terziivanov, MD,PhD,DSc, Professor
Clinic for Therapeutics and Clinical
Pharmacology, Univ Hosp "St. I.Rilsky",
15 D. Nestorov st, 1431 Sofia, Bulgaria
e-mal: terziiv.at.yahoo.com
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)