Hi Shunxin, Interesting project and results! I have a question about the methods: why did you Ln-transform cf-PWV?
Thanks for your question. In our dataset, cf-PWV is skewed distributed. So we ln-transformed it. Ln-transformed cf-PWV is normal distributed.
Well that’s interesting, as I’ve used cf-PWV data of The Maastricht Study twice and it was normally distributed in both instances. Maybe there’s selection bias at play?
It may be selection bias.
Is cf-PWV in your study normally distributed?
Hi Wenjie, nice to see you in the chat!
Either normally or very close to normally distributed (maybe there is selection bias at play in my study, as based on my determinant all individuals who use insulin were excluded).
Regardless, linear regression does not require the determinant and outcome to be normally distributed. Only the residuals need to be normally distributed.
The interpretability of the results is much less complex if you don’t have to transform cf-PWV.
Hi Yuri, agree. How do you think whether the distribution would affect the results via cook’s distance?
Well, that’s a bit difficult to say without looking at the data. If there are outliers detected in the analysis, I would look at the results including and excluding the outliers, rather than transforming the data.
cf-PWV is skewed distributed, ln-transformed cf- PWV is normally distributed.