And I thought it was just behavioural ecologists....
Oh dear [insert appropriate mythical being].
On the Ecological Society of America's list server there has just been an ongoing discussion, which seems to have now run its course, about the nature of the science and the roll of statistics. I might be mischaracterising some of the participants somewhat, but it sure seemed like some people objected to the increasing roll of statistics in ecology.
I'm having trouble parsing this statement. Mainly because the advice I'd expect to get from a statistician would be about things like making sure that I have enough controls, thoughts about sample sizes so that I actually have the power to detect any effects. If you just want to tell a story, by all means go out there and observe nature and write it up. If you want to show that something is causative, then well designed experimental studies, or appropriate statistical analysis of observation data are required. Anything else and you're just pissing into the wind. You're wasting your time and resources.
There was an interesting disscussion about Hamerstrom Science
On the surface it sounds like a good idea. On the other hand there is the minor issue of logistics. Resources are limited, so it's not feasible to just go out and collect all possible data. We need a way to work out what data we should collect, given that we will not be able to collect it all.
So to sum up, in an ideal world statistical analyses would be unnecessary. Unfortunately we don't live in a perfect world so we need those statistical tools to show cause and effect. Even if that's as simple a making sure that our experiments are well designed, and carefully thought out. Whining about the role of statistics is not going to change anything. Regardless of your particular definitions of science we still need to be able to determine if our observations match our hypotheses, or if they aren't different from the appropriate null model.
On the Ecological Society of America's list server there has just been an ongoing discussion, which seems to have now run its course, about the nature of the science and the roll of statistics. I might be mischaracterising some of the participants somewhat, but it sure seemed like some people objected to the increasing roll of statistics in ecology.
But as I have argued several times before, we need to be very wary of letting statisticians tell us how to do science. And once again I want to point out that when I refer to modelling I refer to representations of how the system works, and not to statistical models, which are just arbitrary fits to the data.
I'm having trouble parsing this statement. Mainly because the advice I'd expect to get from a statistician would be about things like making sure that I have enough controls, thoughts about sample sizes so that I actually have the power to detect any effects. If you just want to tell a story, by all means go out there and observe nature and write it up. If you want to show that something is causative, then well designed experimental studies, or appropriate statistical analysis of observation data are required. Anything else and you're just pissing into the wind. You're wasting your time and resources.
There was an interesting disscussion about Hamerstrom Science
Hamerstrom Science, as described by Dr. Joe Schmutz in his published letter in a past issue of the Journal of Raptor Research, emphasized uninhibited, massive gathering of data over long periods of time. There was no formula for the gathering of the data other than to provide all the knowledge that could be gathered on a species or habitat under study, with the goal of gaining personal knowledge until patterns of nature would be identified and explained.
On the surface it sounds like a good idea. On the other hand there is the minor issue of logistics. Resources are limited, so it's not feasible to just go out and collect all possible data. We need a way to work out what data we should collect, given that we will not be able to collect it all.
So to sum up, in an ideal world statistical analyses would be unnecessary. Unfortunately we don't live in a perfect world so we need those statistical tools to show cause and effect. Even if that's as simple a making sure that our experiments are well designed, and carefully thought out. Whining about the role of statistics is not going to change anything. Regardless of your particular definitions of science we still need to be able to determine if our observations match our hypotheses, or if they aren't different from the appropriate null model.
posted by Darkling at 17:08
0 Comments:
Post a Comment
<< Home