Last year at a statistical meeting I met Seth Roberts, a psychology professor at Berkeley and at a Tsinghua University in Beijing. I had been following his blog for a while, and found him to be a fascinating writer, so I was looking forward to the talk. He delivered one of the most interesting hours of entertainment you can imagine at a statistics conference.
Seth has devoted countless time and energy to self-measurement and understanding. He has been monitoring and recording data about himself for many years, mostly to learn about how his daily practices and his diet impact his life. He has discovered many interesting things about himself this way:
- The best way he found to lose weight is to eat “flavorless calories” between meals – calorie dense but bland things like extra light olive oil. He found that this lowered his ‘set point’ and caused him to get full faster when eating, ultimately lowering weight. He wrote a book about it, called the Shangri-La Diet. It is a bizarre concept, a diet which does not actually restrict eating, but instead encourages eating empty calories between meals.
- He discovered that eating half a stick of butter per day improves his brain function, as measured on a standardized test of simple arithmetic questions that he takes every day. (He has made R code available for his daily math test)
- He has found that eating pork fat helps him sleep better.
- Fermented foods are essential for good health and blames the lack of fermented foods in American cooking as a reason for our general poor health and obesity.
- Eating breakfast makes him wake up earlier – skip breakfast and sleep longer and better.
- Sunlight exposure during the day helps him sleep better
- Standing on one leg until exhausted has a huge impact in helping him sleep better (you might have guessed that he has sleep issues)
His paper covers a lot of the above topics. These ideas are all somewhat derided by the mainstream as not having any real ‘science’ behind them.
Which brings us back to the talk at the statistics conference. He covered a lot of the above material and discussed how many of the discoveries were stumbled upon accidentally. But by taking good records of his diet and his sleep and his emotions he was able, after many years, to look at the data and see the correlations. He went beyond this and claimed that he was sure that these results were not just valid for him, but for humans in general. Butter does not make him sleep better. Butter makes humans sleep better. There is no way that he could see such a strong effect and have this not be due to some biological process. He didnt know what that process was, but he was 100% sure it was true.
That was the fun part. He said this to a room of biostatisticians who are ingrained with the dogma of the scientific method and well-designed clinical trials being the only way to really understand biological effects like these. Seth was telling them to take their sample sizes and fancy randomization and shove it. He knows these things to be true because they happened to him.
The incredulous response of the biostat crowd turned to indignation and rejection pretty fast. But he kept pushing, saying isnt it possible that I know my body well enough to know that this must be true? And the effect is so strong that it cannot only be true for Seth Roberts’ body? Don’t human beings have core similarities?
More importantly, he had an important message for science: Imagine if thousands of people did experiments on themselves and kept copious notes on themselves. Imagine if all of these notes were shared and the community allowed to analyze the data. Imagine how much we could learn about diseases like cancer by running thousands of concurrent experiments around the world. Today, running a proper clinical trial requires doctors, biostatistians, and federal regulators, and costs millions of dollars over several years to do one experiment. Seth estimates he does about 200 experiments on himself a year, and estimates the cost at a few hundred dollars. Which is the more efficient model?
The fury in the room was palpable. Most dismissed him out of hand. But is it so crazy? We are starting to see patient led, crowdsourced learning through clever sites such as Patients Like Me. Others say this movement of ‘citizen science’ is the wave of the future. As a statistician, I know that we will still need to use good practices to collect and analyze data, understand variance, correct for biases, and assess significance. But perhaps we need to be a bit less dogmatic and listen more to people like Seth.