dissertation clarification

[Hudson]

Hi everyone
My dissertation topic is ''analysis of the effect of regression to the mean in repeated blood pressure measurements''
My question is in this study there is no treatment involved, I am asking if there is anyone who knows the formula on how I can estimate the effect of the regression to the mean when no treatment is involved.
Waiting to hear from you..
Hudson

 

[katxt]

You measure the BP of a group of people and then, without treatment and under similar conditions, you measure them again. Perhaps unexpectedly, you find that on average, those with the highest BPs the first time tend to have reduced BP levels the second time and those with the lowest BP the first time tend to have slightly increased BP the second time on average. In other words, on average the BPs after tend to be close to the mean BP than the BPs before. But note, this has nothing to do with physiology - it is purely a statistical artifact. (Perhaps even more unexpectedly, those with the highest BPs the second time tended to have BPs a little lower the first time, and vice versa.)
If you plot a graph of BP after up, and BP before across (or vice versa) you get the expected upward trend, and if conditions are the same you would expect the slope of the best fit line to be 1. However, it will be less than that. In this situation the slope is a measure of the correlation between the before and the after BPs and would be an estimate of the effect of the regression to the mean.

 

[Hudson]

You measure the BP of a group of people and then, without treatment and under similar conditions, you measure them again. Perhaps unexpectedly, you find that on average, those with the highest BPs the first time tend to have reduced BP levels the second time and those with the lowest BP the first time tend to have slightly increased BP the second time on average. In other words, on average the BPs after tend to be close to the mean BP than the BPs before. But note, this has nothing to do with physiology - it is purely a statistical artifact. (Perhaps even more unexpectedly, those with the highest BPs the second time tended to have BPs a little lower the first time, and vice versa.)
If you plot a graph of BP after up, and BP before across (or vice versa) you get the expected upward trend, and if conditions are the same you would expect the slope of the best fit line to be 1. However, it will be less than that. In this situation the slope is a measure of the correlation between the before and the after BPs and would be an estimate of the effect of the regression to the mean.

Thanks Katxt,
How do estimate the slope of best fit in STATA software? Do I run a linear regression? How do I estimate effect of regression to the mean in STATA?Please explain more.
I value and appreciate your help.
Hudson.

 

[katxt]

The best fit line is just a linear regression. You can do both before vs after, or after vs before. The two slopes should be much the same and less than 1. To get the best value, the correlation between the two sets is sort of an average of the two. If there was perfect matching between before and after, the slope and the correlation would be 1. Let's say the correlation and the two slopes are, in fact, about 0.8. Then the average means have regressed towards the mean by about 1-0.8 or about 20%. kat