Any movement that lacks explicit acknowledgement of statistical significance is worrisome to me. I think for many -- myself included -- it is temptingly easy to attribute causation or project a bias onto data collected like this (eg "I'm in a bad mood today because I only slept for 5 and a half hours last night").
I would be interested in analyzing the data after I had a large pool and a decent idea of shape, center, and spread, but I also don't trust myself to wait that long.
> Any movement that lacks explicit acknowledgement of statistical significance is worrisome to me.
I'm not necessarily too worried about that. Statistical significance is the wrong concept for QS and its use is essentially cargo cult statistics.
Leaving aside the profound conceptual and applied problems with null-hypothesis testing ( http://lesswrong.com/lw/g13/against_nhst/ ), QS is much closer to cost-benefit analysis where effect sizes and costs are the critical variables, not alpha. We don't care about testing some intervention and not making the completely arbitrary cutoff of 0.05 (which doesn't mean anything about the truth of the hypothesis in the first place)! We care whether the intervention make a large impact on the variable in question and how expensive the intervention was; if, say, the intervention is an expensive supplement that costs hundreds of dollars a year, we want a higher burden of proof than if the intervention is something free (like taking your vitamin D supplement in the morning rather than evening) or something we should be doing anyway (like exercise).
Far* more worrisome than QS's failure to run t-tests and ritually chant 'we calculate a p-value of <0.05 therefore we reject the null hypothesis of no difference' is the pervasive publication bias (who reports failed experiments?), the absence of blinding even where quite easy leading to severe placebo effects (many supplements), tiny sample sizes, and dodgy data collection (selection bias).
I'm torn, because QS is something I'm interested in and something I think many could benefit from, but at the same time this article reads like an advertisement, especially since it suspiciously lacks mention of well known alternatives. You could argue that they only mention what they think of as "the best", but I would argue that most of these tools put far too much of your information into other peoples' hands. As an alternative, at least for biometrics, consider the following list of bluetooth enabled devices which can be used with software that doesn't upload your data to someone else's servers:
I'd also have to recommend org-mode in Emacs, at least for time tracking. I know it's not "mobile", but I use it on my phone to track all sorts of things, and since the format is plain text, I can slice and dice in any language I want, no third parties needed.
Narrato are building an open platform for the quantified self. They currently only have one sample app called Narrato journal but it might be one to keep an eye on:
https://www.narrato.co/
Interesting. I just started trying out BeeMinder myself, which seems like a good tool to both track things about yourself, and commit to changing your behaviour.
They kept mentioning the "quantified self" and I didn't know what they were talking about.
I've found that I'm actually a lot happier not tracking all this stuff about myself. I was tracking my sleep, my running, my eating, my stress, my time, etc. Then a few months ago decided that I needed to focus more on living life and less on just checking in to it.
I think that finding your optimal level of data tracking, or even ramping it down after building a habit is key to these sorts of things. It's a cost/benefit tradeoff, really. The thing is that the cost side is being changed by the fact that much of the data collection can be automated these days, and hopefully the analysis side can be automated as well, so that you don't have to focus as much on the numbers.
I find it useful if it's mostly automated. I use a basis band for sleep & activity, and a fitbit aria scale for weight. It being automated doesn't stress me out and it's useful to have a history to see trends.
I also find it fun to start some runkeeper app when I go on bike rides, and that only takes about a minute or two before I start it. I never actually look at the history of my cardio workouts, but it's fun to have the stats while I'm biking, and to see a map of where I went after I'm done.
My problem is that the tool I used isn't automatic. I just got too tired of pricking my skin for glucose tracking, noting down my weight, and resetting my pedometer everyday.
May be it was because you were tracking too many things? I think it is like tweeting/taking photos etc - done sparingly, it might be really useful (especially when tracking only important things), doing it too much might feel like a chore.
The old school version of this is called the Personal Software Process, first described in A Discipline for Software Engineering. It was sometimes used as a punchline by early agile advocacy.
What would be nice is to find a decent site/app for tracking food intake. That is one area that I would love to track but the current landscape of apps for it are lacking.
MyFitnessPal's UI is pretty bad but seems to have decent data if you can find it.
Fitbit's UI is nice but the food database and lookup is atrocious.
I absolutely agree. It is absurd that the two most popular Fitness/Diet apps are so poorly designed, especially given the open nature of the ecosystem. Users, developers: why?!
AskMeEvery.com is blocked by Trend Micro for me saying "Dangerous Verified fraudulent page or threat source." I assume it is a false positive but I wanted to bring it up in case Mark reads this so that he can look into getting Trend to fix it.
Meh. This quantified self business is just the next iteration of herbal remedies, Chinese medicine, and all that other pseudo-scientific mumbo-jumbo out there. You might as well add your astrological sign to all these apps.
This is yet another area where people are attempting to layer technology on top of a problem that does not exist.
Meh. This quantified self business is just the next iteration of herbal remedies, Chinese medicine, and all that other pseudo-scientific mumbo-jumbo out there. You might as well add your astrological sign to all these apps.
Why would you think that? If anything, the self quantified movement is all about thinking scientifically to improve yourself. Gathering data, analyzing, designing experiments, generating hypothesis, and publishing for feedback are all part of being self quantified.
Sure, it will be used for selling a lot of snake oil, like anything related to health or productivity.
But at the same time, we're surrounding ourselves with so many sensors that for the first time in the history of mankind we can really "data mine ourselves". This may bring up interesting discoveries.
In the early days of science some kept strict logs about everything they did with the same goal. Now this can be automated.
What you are missing is that people figuring out useful stats from all these numbers and then improving their lives based on real life in real time. I suggest you try this for 1 day, just keep track of your happiness. Its really interesting and optimizable, as you can find out.
Sure no one is saying quantified will get you from a 1 to a 10 on the happiness scale, but it will definitely help you to learn more about yourself.
[+] [-] apawloski|13 years ago|reply
I would be interested in analyzing the data after I had a large pool and a decent idea of shape, center, and spread, but I also don't trust myself to wait that long.
[+] [-] gwern|13 years ago|reply
I'm not necessarily too worried about that. Statistical significance is the wrong concept for QS and its use is essentially cargo cult statistics.
Leaving aside the profound conceptual and applied problems with null-hypothesis testing ( http://lesswrong.com/lw/g13/against_nhst/ ), QS is much closer to cost-benefit analysis where effect sizes and costs are the critical variables, not alpha. We don't care about testing some intervention and not making the completely arbitrary cutoff of 0.05 (which doesn't mean anything about the truth of the hypothesis in the first place)! We care whether the intervention make a large impact on the variable in question and how expensive the intervention was; if, say, the intervention is an expensive supplement that costs hundreds of dollars a year, we want a higher burden of proof than if the intervention is something free (like taking your vitamin D supplement in the morning rather than evening) or something we should be doing anyway (like exercise).
Far* more worrisome than QS's failure to run t-tests and ritually chant 'we calculate a p-value of <0.05 therefore we reject the null hypothesis of no difference' is the pervasive publication bias (who reports failed experiments?), the absence of blinding even where quite easy leading to severe placebo effects (many supplements), tiny sample sizes, and dodgy data collection (selection bias).
* If you are wondering why anyone would care about my opinion, I've been self-experimenting for years and have a little bit of insight into the matter; see http://www.gwern.net/Zeo http://www.gwern.net/Nootropics and http://www.gwern.net/Weather
[+] [-] npsimons|13 years ago|reply
http://personalheartmonitor.com/sensors.html
I'd also have to recommend org-mode in Emacs, at least for time tracking. I know it's not "mobile", but I use it on my phone to track all sorts of things, and since the format is plain text, I can slice and dice in any language I want, no third parties needed.
[+] [-] joefarish|13 years ago|reply
On the subject of quantified self Stephen Wolfram's blog post 'The Personal Analytics of My Life' (http://blog.stephenwolfram.com/2012/03/the-personal-analytic...) is an interesting read.
[+] [-] edanm|13 years ago|reply
They kept mentioning the "quantified self" and I didn't know what they were talking about.
[+] [-] bhousel|13 years ago|reply
[+] [-] npsimons|13 years ago|reply
[+] [-] mahyarm|13 years ago|reply
I also find it fun to start some runkeeper app when I go on bike rides, and that only takes about a minute or two before I start it. I never actually look at the history of my cardio workouts, but it's fun to have the stats while I'm biking, and to see a map of where I went after I'm done.
[+] [-] kiba|13 years ago|reply
[+] [-] vijayr|13 years ago|reply
[+] [-] jacques_chester|13 years ago|reply
[+] [-] marmalade|13 years ago|reply
MyFitnessPal's UI is pretty bad but seems to have decent data if you can find it.
Fitbit's UI is nice but the food database and lookup is atrocious.
[+] [-] ksyeung|13 years ago|reply
[+] [-] noneTheHacker|13 years ago|reply
[+] [-] vbl|13 years ago|reply
[+] [-] sheraz|13 years ago|reply
This is yet another area where people are attempting to layer technology on top of a problem that does not exist.
[+] [-] kiba|13 years ago|reply
Why would you think that? If anything, the self quantified movement is all about thinking scientifically to improve yourself. Gathering data, analyzing, designing experiments, generating hypothesis, and publishing for feedback are all part of being self quantified.
[+] [-] bergie|13 years ago|reply
But at the same time, we're surrounding ourselves with so many sensors that for the first time in the history of mankind we can really "data mine ourselves". This may bring up interesting discoveries.
In the early days of science some kept strict logs about everything they did with the same goal. Now this can be automated.
[+] [-] rickdale|13 years ago|reply
Sure no one is saying quantified will get you from a 1 to a 10 on the happiness scale, but it will definitely help you to learn more about yourself.