One of the more surprising things about the sports massage job is that I have been able to speak about all sorts of topics which interest me with clients who are professionals in that field, and hear their thoughts.
This has ranged from talking about money and economics with accountants to healthcare, contraindications and cauda equina with doctors and management consultants.
One highlight has been finding that one of my clients is a high level researcher who is attempting to investigate indolent prostatic cancers to attempt to find a way to differentiate them from fully malignant ones.
But the best bit has been harassing statisticians and economists about linear multivariate regression analysis!
The economists were fairly uniform in thinking that there was no problem with linear only models - one of them even asked "what would you want to model that wouldn't work in a straight line?" and when I tried to introduce my hopeful idea that logarithmic transformations could be used to get a linear model from a non-linear set couldn't understand the point.
So I was thrilled to finally find the poor statistician who actually agreed with me that it was bizarre that non-linearity was not further investigated. She confirmed my fears that there were no good proofs so that proper modelling techniques could be developed, but agreed with me that mapping and transforms would be a good idea. She said she was pleased to work in a field where it appeared that the models they were creating were all truly linear, so there was no issue.
Vindication at last!
Of course, this doesn't change the fact that my dissertation didn't really find much. But I did enjoy getting my stats back!
This has ranged from talking about money and economics with accountants to healthcare, contraindications and cauda equina with doctors and management consultants.
One highlight has been finding that one of my clients is a high level researcher who is attempting to investigate indolent prostatic cancers to attempt to find a way to differentiate them from fully malignant ones.
But the best bit has been harassing statisticians and economists about linear multivariate regression analysis!
The economists were fairly uniform in thinking that there was no problem with linear only models - one of them even asked "what would you want to model that wouldn't work in a straight line?" and when I tried to introduce my hopeful idea that logarithmic transformations could be used to get a linear model from a non-linear set couldn't understand the point.
So I was thrilled to finally find the poor statistician who actually agreed with me that it was bizarre that non-linearity was not further investigated. She confirmed my fears that there were no good proofs so that proper modelling techniques could be developed, but agreed with me that mapping and transforms would be a good idea. She said she was pleased to work in a field where it appeared that the models they were creating were all truly linear, so there was no issue.
Vindication at last!
Of course, this doesn't change the fact that my dissertation didn't really find much. But I did enjoy getting my stats back!