You are viewing a single comment's thread from:
RE: Where Are All The Flying Insects Going? Researchers Observe A 75% Decrease In The Amount Of Flying Insects Over The Past 27 Years
However we can also see a VERY clear trend with regards to the average value and this trend goes down by quite a lot over the years.
As a disclaimer, I accept this conclusions. However, being the devil's advocate, the results seem to be as compatible as being constant over time, when accounting for the error bars.
Note that I am also very amazed by the conclusions of the study that are very scientific and objective!
PS: why Germany?!
In this data the error reported is the standard deviation, perhaps reporting confidence in the mean would be a more appropriate error determinant for this sort of data? (I don't think the authors are (should be?) as concerned with ease of replication study as they are in the trust-ability of their measurements )
Beyond this, the y-axis scale of the plot is non linear, which masks the magnitude of the change, it would appear more obvious with a different scale.
All of that said, the data is really noisy but what can one expect from catching bugs in nets? ;)
I think they did a poor job of presenting this data. You definitely want to use standard error over standard deviation in this case, and the log scale minimizes the appearance of the effect.
I agree. You can check my answer above (sorry, I only noted your comment after having answered to @justtryme90) :)
That is right. In noticed the log scale which is really not a good choice (since the range covers two orders of magnitude, I guess presenting the results under both forms (log and linear ) could have been better.
But even, I think a flat curve would work as well. Having compared both options and given the outcome as the confidence level of the trend would have been better...
Anyways, I am always biased by my physics background where anything can be properly (well, most of the time) quantified. As you said elsewhere, I should learn thinking as a biologist (or at least stop commenting biology works as a physicist :p ).
Nah, you should keep your high expectations that physics holds. Wanting more proof from data or to be better convinced is never a bad thing.
I should pass the message in some way. Jut to let you know, at each interview of a biophysicist, this is always the same question I ask: error bars, confidence level, etc... (yes, biophysics is part of theoretical physics so that I may interview those guys :p )
Biophysics is not technically a part of theoretical physics by terminology. Much of the work I do falls under that term for pharma as I define molecular interaction strength. Biophysics in that context falls under biochemistry ironically.
Honestly a good understanding of error propagation and when to apply a particular analysis is a must in any scientific discipline. It is also something that a lot of people don't understand. You would be surprised (well YOU personally probably wouldn't be surprised) how few people understand the difference between simple concepts like standard deviation and standard error for instance.
It depends how you define theoretical physics. In France, it really includes the theory side of particle physics, astrophysics and cosmology, formal stuff like string theories or mathematical physics, condensed matter, statistical physics and finally areas like turbulence and biophysics. At the end of the day, it turns out that theoretical physics is kind of multidisciplinary.
it is very rare to have good statistics courses in standard curriculums. This is where one should start to hope for a change. Even in my field, sometimes... (remember the Madala story I wrote about some time ago... we can chat about it on the steemit chat if you want and I will give you more details about the statistics procedure and its weak spots; which does not prevent to recognize that the Madala exercise is interesting per se).