Recent papers you have read

[Ex-User (14663)]

Post a paper you've recently read, your thoughts on it, what it was about etc. Perhaps even a critique of it?

I'm sure this thread will be a great hit.

I will be back when I have read a paper.

 

[Hadoblado]

Harrington, R. & Loffrendo, D. (2001). The relationship between life satisfaction, self-consciousness, and the Myer-Briggs Type Inventory dimensions. The Journal of Psychology, 135(4), 439-450.

A few things:
Sample was college students, but mean age was 31. That sounds like outliers are skewing the numbers a lot, and this is not represented in the report. I'm a mature age student and I'm only 28. The average first year uni student is like 19 or something?

The vast majority of the participants were female (80%ish). There are sex effects in MBTI, yet this wasn't a controlled variable? I understand that university research is typically made on convenience samples, but it just feels like an unnecessary muddying of results. I'd personally just do the research on women if I'm having difficulty matching numbers.

They used the dichotomy 'dimensions' instead of functions or types. I'm not great with this stuff, but I thought the dichotomies were supposed to be shallow descriptors? They still found a whole bunch of effects at the p = <.001 level so what do I know right?

I didn't really rate the paper. I read it cos I was on hold and it was the first thing that came up in the search engine. It wasn't really exploring anything interesting, and it seemed a little sloppy.

 

[Cogitant]

Spent 3h on the phone to my best friend who is a psychiatric nurse, we've been discussing some of the patients where she works (not named due to confidentiality), their symptoms, and debating psychosis; the cause, effect and treatments for it.
[FONT=&quot]The debate was a long one, (INTP - INFP debates are really deep and thorough). I'm interested in the scientific method and treatment via chemical rehabilitation, she thinks that chemical treatments are not the way, talking and alternative methods are superior, and society should be more tolerant of the mentally ill (she's a spiritual and caring soul, has a point tbf).
[/FONT]
Anyhow, this led me to do a spot of research regarding the science behind psychosis:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2838993/#R2

The article highlights the fact that perhaps dopamine imbalance alone is not responsible for psychosis, but also glutamate.

[FONT=&quot]"It raises the question of whether or not the development of a primarily glutamatergic substance might be a promising route for a new antipsychotic treatment strategy."

[/FONT]Seems as if anti-psychotic medicines are greatly in need of revision (also in some cases, severe side effects of common medicines such as chlorpromazine hydrochloride can outweigh the positives).
[FONT=&quot]
The document also got me thinking about the genetic factors involved.
It could be helpful to understand if one has a genetic risk of developing psychotic illness, so that one can learn to avoid trigger situations and/or substances.

In all, a well written, high quality document.
Gave plenty of food for thought, and has piqued my interest in neurochemistry.

-I believe that talking and alternative therapies, as my friend suggested, could be helpful before, after and between psychotic episodes. It could aid in treating the cause and not just the symptoms.
During psychosis, however, anti-psychotic medicine would be strongly advisable.


[/FONT]

 

[Ex-User (14663)]

Probably on the esoteric end of subjects so far, but I read Analysing financial returns by using regression models based on non-symmetric stable distributions by Lambert, Lindsey (1999).

Based on the title, it's about regressions involving stock returns, but really, it's applicable to anything. Whenever you see a piece of research providing, say, correlations and estimations of means, but doesn't mention anything about the soundness of the assumptions about the underlying distribution, those estimates become very dubious. For example, if the sample correlation is to be interpreted as an estimator of the population correlation, you have to assume the distribution that produces the sample even has a second moment (i.e. variance). For a lot of things in the world that assumption is false – for example for income. Large deviations in those cases are typically deemed "outliers" by researchers while really, it's just that the underlying distribution is Stable and produces large deviations by virtue of being skewed and/or leptokurtic.

The results of the example have major implications for most of traditional statistical modelling. In the history of statistics, how many spuriously significant differences in means have been detected when, in fact, only the shape of the distribution (dispersion or variance, skewness or the thickness of tails) was changing? What indeed is the appropriate measure of difference in `location' for a given problem? For example, are we interested in how the mean or the most probable response changes? All too often in statistics, means are calculated and studied without any reference to the distribution from which they are derived. But we see that a mean, or location parameter, cannot be correctly interpreted (the mean may not even exist) if the distribution from which it arises is not specified, especially if that distribution may possibly be skewed.

The main contribution of this paper is a maximum-likelihood procedure for doing regression and simultaneously fitting the parameters of a stable distribution. It seems to work reasonably well, but based on some tests I have run it is extremely sensitive to initial conditions and often gets stuck doing numerical integration for ages. I intend to write an improved version of this procedure.

 

[Turnevies]

Is the statement of Murphy's law valid? http://arxiv.org/pdf/1508.07291.pdf

Spoiler: no, according to the authors. Contains formal logic, measure theory, thermodynamics and calculus of variations. I mostly get their line of reasoning, but I'm not able to discuss the validness of the approach entirely.

 

[Ex-User (14663)]

Is the statement of Murphy's law valid? http://arxiv.org/pdf/1508.07291.pdf

Spoiler: no, according to the authors. Contains formal logic, measure theory, thermodynamics and calculus of variations. I mostly get their line of reasoning, but I'm not able to discuss the validness of the approach entirely.

That's funny. About 2 years ago I asked this question on math.stackexchange. Ultimately, I guess from a purely mathematical point of view you can come to any conclusion you want, depending on the assumptions. So it's definitely more interesting to look at it from a physical point of view.

 

[Turnevies]

That's funny. About 2 years ago I asked this question on math.stackexchange. Ultimately, I guess from a purely mathematical point of view you can come to any conclusion you want, depending on the assumptions. So it's definitely more interesting to look at it from a physical point of view.

 

[Ex-User (14663)]

B. Mandelbrot – The variation of certain Speculative Prices, 1963

Mandelbrot shows in this paper the inadequacy of the Normal distribution as a model for financial returns, and proposes an alternative model – the stable distribution.

It's quite an experience reading this paper. Mandelbrot wrote it in 1963 and it just shows how severe the decline in the quality of academic writing has been since then.

In a surprisingly lively argument following Cauchy's 1853 paper, J. Bienaymé stressed that a method based upon the minimization of the sum of squares of sample deviations cannot be reasonably used if the expected value of this sum is known to be infinite.

It is hilarious to read this. In 1853 they understood that least-squares doesn't work when the variance is infinite, yet this fact somehow escapes virtually all modern social-scientists, economists and people who do research in finance (I actually tried to explain this exact fact to my former employer, but to no avail).

All in all a masterpiece of a paper.

 

[Ex-User (14663)]

Billings, Darse, et al. "Approximating game-theoretic optimal strategies for full-scale poker." IJCAI. 2003.

Interesting in the sense that it provides a computable method for obtaining Nash equilibria for "full-scale" poker. This is basically done by grouping a bunch of cases together to reduce the number of cases to optimize over. Although, currently I don't understand how it is possible to have a Nash equilibrium against an arbitrary opponent and what assumptions they have made in that regard. Have to look into some of the references.