Dear All
I am a PhD student in Medical Gerontology and I would be very grateful for some
help regarding a variable transformation problem.
I have a continuous normally distributed variable (the risk factor) that I am
trying to correlate with a continuous variable which has severe positive
skewness (the outcome).
I am using SPSS 16.0.
The cubic fit method provides the best fit for the correlation between these two
variables (RSqCubic = 0.6).
Which means that as the risk factor increases linearly, the outcome increases
exponentially, especially for the highest levels of the risk factor. In the
correlation scatterplot there is almost a flat line until about mid values of
the risk factor, followed by a sharp bend and a huge exponential increase.
I would like to transform the risk factor so that it reflects linear increases
in the outcome. What transformation should I use? What would be the formula I
need to enter in the "Compute Variable" menu?
Thank you very much in advance for your help
Roman Romero
|