I thought we were talking about the reconstruction filter.
Can you please post a link to his comments/studies on this? I am always interested in he's opinions/findings.
I have no clue where he has posted this publicly. It was in a private conversation. We were disagreeing on the advantages/disadvantages of very steep filters and merit of different approaches to phase linearization. He was making his case that he thinks both are a waste of time.
he has cited research to me before that was never published. I’ve asked his team about this before and the answer I got was that most groups and most researchers are uninterested in publishing negative findings. As a researcher myself, this is actually a know problem with the field. It’s sometimes known as the file drawer effect. Basically, non-significant findings go unpublished and their “information” doesn’t go into the collective knowledge on a topic. Only the positive studies.
this is now totally off topic. However I can’t stop myself since I suspect most people don’t fully understand how research publications work, as most don’t conduct research that ever needs to be published.
there is a known bias amongst academic journals toward positive findings. Studies that fail to find anything are rarely published by the journals unless there is something really compelling in the study. This is more true today than in the past, so we find more negative published studies in the past too. This creates a systematic bias in the research base.
researchers also don’t tend to submit negative findings (I.e. those for which you failed to reject the null hypothesis or even those with findings contrary to your hypothesis) because researchers are unlikely to make tenure based on a body of failed research. Again, this introduced a bias. If we think in Bayesian terms, our priors don’t accurately reflect our true universe. There is now lost knowledge.
we can see this as some grand indictment of the scientific community. It is in fact a problem. But it isn’t so simple either. Remember the common research concept, the absence of evidence is not evidence of absence. This is more true than most realize.
Statistics, used in these studies, are both wonderful tools and horrifically misleading. In the hands of the inexperienced and poorly educated (on the topic of statistics) they get horribly misused. A hypothesis test, as used in these studies, is trying to test just one main hypothesis. But we aren’t. We are really testing 100’s or 1000’s of hypothesis in this one experiment. We only care about one, but the others are there. These are all the confounding factors in the study that could explain a change (or lack of change) in the estimate. In the absence of a lot of knowledge about the problem being tested and a sound theory for how it all works, both positive and negative findings are actually suspect.
mad a result of these statistical complexities, a negative findings can actually be explained by many factors. One of which is that no difference exists. However, we can’t say that because we don’t have enough information to know that. No matter how good the study seemed. Failing to find something (like an audible difference from linear phase filters) is not a finding itself. Hence why journals normally don’t publish them. Hence why researchers normally don’t try to publish them. Hence why institutions normally don’t promote researchers or give tenure when publishing large bodies of non-evidence. Remember, a failure to reject the null hypothesis is not a finding. It’s not evidence. It’s a non-finding and this non-evidence.
However there is always this distinct possibility that you have disproven a theory and this we have learned something important. Because we can’t prove the theory wrong (we can only fail to prove it right repeatedly) we end up with an academic bias.
because of type 1-2, M, and S errors, and because a lot of researchers are actually quite lousy staticians, simply publishing all null findings would be just as problematic. A lot of studies fail because the researchers kind of suck at their job, for lack of a better way of putting it.
all to say, I can’t find any published literature on the inaudibility of lots of things. When I talk to the experts about why they didn’t publish the null findings, this is essentially what they tell me.
But I admit -- don't know the inner-workings of the AKM4458.
