“Scientists from Dana-Farber Cancer Institute, Brigham and Women’s Hospital, and the Harvard School of Public Health came to this conclusion after analyzing data on nearly 120,000 people collected over 30 years.”
“The most obvious benefit was a reduction of 29 percent in deaths from heart disease—the major killer of people in America. But we also saw a significant reduction − 11% - in the risk of dying from cancer.”
The researchers point out that the study was not designed to examine cause and effect and so cannot conclude that eating more nuts causes people to live longer.
Indeed, the study consists only of observational data, not interventional, so what causal conclusions could be drawn from it?
Sorry, there are two separate issues: the data itself (which is a big dataset where they following a big set of nurses for many years, and recorded lots of things about them), and how the data could be used to maybe get causal conclusions.
Plenty of folks at Harvard (e.g. Miguel Hernan, Jamie Robins) used this data in a sensible way to account for confounding (naturally their results are relatively low on the ‘hierarchy of evidence’, but still!) Trying to draw causal conclusions from observational data is 95% of modern causal inference!
Eating a handful of nuts a day.
http://www.medicalnewstoday.com/articles/269206.php
But:
Indeed, the study consists only of observational data, not interventional, so what causal conclusions could be drawn from it?
You act like people never did a valid causal analysis of the data in the Nurses’ health study.
I know I overstated things. There are such things as natural experiments, having some causal information already, etc.
I’m not familiar with the Nurses’ health study, and a quick google only turns up its conclusions. What methods did they use?
Sorry, there are two separate issues: the data itself (which is a big dataset where they following a big set of nurses for many years, and recorded lots of things about them), and how the data could be used to maybe get causal conclusions.
Plenty of folks at Harvard (e.g. Miguel Hernan, Jamie Robins) used this data in a sensible way to account for confounding (naturally their results are relatively low on the ‘hierarchy of evidence’, but still!) Trying to draw causal conclusions from observational data is 95% of modern causal inference!