Disclaimer: All the views here are merely based on observations and the authors own experience.
Biology: learning how systems work by breaking them.
What can we learn from biologists?
Ask important and fundamental questions about the real world.
For example why do vaccinations work?
Come up with interesting and innovative ideas and experiments.
They tend to be more in the real world (although that's debatable).
All data is not equal, some datasets are worse quality than others.
We statisticians have metrics like the CV to measure the signal to noise ratio, but the problem is we don't really understand what is the signal in the first place.
We also only know about what is actually measured in the experiment and not about confounders.
What biologists can learn from us?
Rigour. Naming conventions and standards are important dammit.
Being organised! Store data preferably in text files so that it is machine readable, easily searchable without proprietary software etc
Efficiency!
Objectivity, not let our beliefs cloud our judgement, influence our analysis/experiments.
This goes for both us but "don't let data get in the way of a good story" is not science.
All datasets are important. Small datasets do contain useful information which can sometimes be exploited when joined on other datasets.
Statistical common sense often not shared with biologists:
Winners curse or regression to the mean
Multiple testing
Simpson's paradox
Biology: learning how systems work by breaking them.
What can we learn from biologists?
Ask important and fundamental questions about the real world.
For example why do vaccinations work?
Come up with interesting and innovative ideas and experiments.
They tend to be more in the real world (although that's debatable).
All data is not equal, some datasets are worse quality than others.
We statisticians have metrics like the CV to measure the signal to noise ratio, but the problem is we don't really understand what is the signal in the first place.
We also only know about what is actually measured in the experiment and not about confounders.
What biologists can learn from us?
Rigour. Naming conventions and standards are important dammit.
Being organised! Store data preferably in text files so that it is machine readable, easily searchable without proprietary software etc
Efficiency!
Objectivity, not let our beliefs cloud our judgement, influence our analysis/experiments.
This goes for both us but "don't let data get in the way of a good story" is not science.
All datasets are important. Small datasets do contain useful information which can sometimes be exploited when joined on other datasets.
Statistical common sense often not shared with biologists:
Winners curse or regression to the mean
Multiple testing
Simpson's paradox
No comments:
Post a Comment