We managed to reduce performance on any number of tests to essentially a single number, g, together with a couple more for domain-specific skill. We managed to reduce the huge variation in personalities to five numbers, the OCEAN dimensions. I even recall reading that there is quite some correlation between those five numbers and that they might be reduced to a single one but I can’t find the source any more.
Can we construct a whole host of other, similar numbers, like “math skills” and thus measure the impact of education and aging?
Another number I have in mind is, can we construct three numbers general health gh, mental health mh and physical health ph, and measure their correlations? I have the vague observation that medical issues tend to cluster, that is people with mental issues tend to not only exhibit any one of ADHD, depression, OCD and so on, but more than one of them. Similarly I have the impression that people tend to complain of many physical symptoms at once.
I seem to recall that BMI and/or WHR tend to be excellent predictors of physical health. Together with a couple of more measures these predictions can further be improved. The advantage of having a single number would be for research purposes on population health and it is easier to have a single mumber for personal assesment.
Not quite reduce. We managed to develop certain approximations which, albeit crude, work sufficiently well for some purposes. Of course, not all purposes.
I seem to recall that BMI and/or WHR tend to be excellent predictors of physical health.
I seem to recall they tend not. In particular, BMI is a flawed indicator as it has a pronounced bias for short and tall people.
these predictions can further be improved
Which “these predictions”—what are you forecasting?
A high pressure is a good predictor for someone being unhealthy. On the other hand statins that reduce blood pressure don’t provide the returns that people hoped for.
Goodhard’s law applies very much.
Before dying with a heart attack Seth Roberts had a year where he improvement on the score that’s the best predictor for heart attacks, while most people don’t improve on the score as they age.
Using metrics like BMI and WHR seems to me very primitive. We should have no problem running a 3D scan of the whole body. I would estimate that obesitey[3D scan + complex algorithm] is a much better metric than obesity[BMI], obseity[WHR] or obesitey[BMI/WHR].
That’s to be further improved by not only going for the visible light spectrum but adding infrared to get information about temperature. And you can follow it up by giving the person a west with hundreds of electrodes and measuring the conductance.
As quantified self devices get cheaper it will also be possible to use their data to develop new metrics. A nursing home could decide to give every member a device that tracks heart rate 24⁄7. After a few years time the can give the data to some university bioinformatics folks who try to get good prediction algorithms.
Can we construct a whole host of other, similar numbers, like “math skills” and thus measure the impact of education and aging?
Math skills can mean multiple things to different people. Some people take it to mean the ability to calculate 34*61 in a short amount of time and without mistakes. Other people take it to mean doing mathematical proofs.
We might even find something more sophisticated than fat percentage. Not all fat people are ill/heading towards illness. Not all thin people are healthy.
We managed to reduce performance on any number of tests to essentially a single number, g, together with a couple more for domain-specific skill. We managed to reduce the huge variation in personalities to five numbers, the OCEAN dimensions. I even recall reading that there is quite some correlation between those five numbers and that they might be reduced to a single one but I can’t find the source any more.
Can we construct a whole host of other, similar numbers, like “math skills” and thus measure the impact of education and aging?
Another number I have in mind is, can we construct three numbers general health gh, mental health mh and physical health ph, and measure their correlations? I have the vague observation that medical issues tend to cluster, that is people with mental issues tend to not only exhibit any one of ADHD, depression, OCD and so on, but more than one of them. Similarly I have the impression that people tend to complain of many physical symptoms at once.
I seem to recall that BMI and/or WHR tend to be excellent predictors of physical health. Together with a couple of more measures these predictions can further be improved. The advantage of having a single number would be for research purposes on population health and it is easier to have a single mumber for personal assesment.
Not quite reduce. We managed to develop certain approximations which, albeit crude, work sufficiently well for some purposes. Of course, not all purposes.
I seem to recall they tend not. In particular, BMI is a flawed indicator as it has a pronounced bias for short and tall people.
Which “these predictions”—what are you forecasting?
And muscular people. What’s wrong with WHR?
A high pressure is a good predictor for someone being unhealthy. On the other hand statins that reduce blood pressure don’t provide the returns that people hoped for.
Goodhard’s law applies very much.
Before dying with a heart attack Seth Roberts had a year where he improvement on the score that’s the best predictor for heart attacks, while most people don’t improve on the score as they age.
Using metrics like BMI and WHR seems to me very primitive. We should have no problem running a 3D scan of the whole body. I would estimate that obesitey[3D scan + complex algorithm] is a much better metric than obesity[BMI], obseity[WHR] or obesitey[BMI/WHR].
That’s to be further improved by not only going for the visible light spectrum but adding infrared to get information about temperature. And you can follow it up by giving the person a west with hundreds of electrodes and measuring the conductance.
The tricoder xprice is also interesting.
As quantified self devices get cheaper it will also be possible to use their data to develop new metrics. A nursing home could decide to give every member a device that tracks heart rate 24⁄7. After a few years time the can give the data to some university bioinformatics folks who try to get good prediction algorithms.
Math skills can mean multiple things to different people. Some people take it to mean the ability to calculate 34*61 in a short amount of time and without mistakes. Other people take it to mean doing mathematical proofs.
We might even find something more sophisticated than fat percentage. Not all fat people are ill/heading towards illness. Not all thin people are healthy.
Accumulation of fat to vital organs like the liver could be a better predictor. Fatty liver can be diagnosed via ultrasound, which is cheap.
Being fat is a risk even if you get sick for other reasons. Rehabilitation suffers.
Cite?
Fatty liver predicts the risk for cardiovascular events in middle-aged population: a population-based cohort study
Obesity and Inpatient Rehabilitation Outcomes Following Knee Arthroplasty: A Multicenter Study
Yes, we have to try many different metrics and see which ones work best and for what purposes.