-
New Forum Software Installed (Still a few issues but reopening the forum) More details.
Treating patients suffering from myalgic encephalopathy/chronic fatigue syndrome (ME/CFS) with sodium dichloroacetate, Comhaire 2018
- Thread starter Indigophoton
- Start date
There seems to be some confusion. The "total FSS" is calculated by adding up the score of each separate item (e.g. the sum is 58). This sum is divided by 63 (e.g. sum= 58, divide by 63= 0.92). Now multiply this by 7 (e.g. 0.92 x 7 = 6.44). The value 6.44 is the total (or average) FSS. Introduce this alue in the formula. The result is logit(p). Next convert logit(p) to (p).
adambeyoncelowe
Senior Member (Voting Rights)
This gives me exactly the same number as Trish's way.
55/63 = 0.8730.
0.8730 x 7 = 6.11111...
55/9 = 6.11111... too.
Feeding that number back into the formula gives me:
Logit(p) = 7.09 + [2.27 * 7] + [3.39 * 7] - [2.57 * 5] - [1.97 * 5] - [2.52 * 6.1] = 8.64.
p = e^8.64 / 1 + e^8.64 = 0.9998
However, I'm terrible at maths!
But there is a serious question as to whether this is a valid operation. The FSS scores are basically just answers to a random set of questions vaguely about fatigue. Is it valid to add them up and average? What is the meaning?
I also don't get the rational for using the logistic function and why the claim for it being a probability?
Woolie
Senior Member
I don't think there's any convincing evidence that psychological stress changes your DNA. You may be overextrapolating from studies of mice that have been placed under extreme physical stress (deprived of food and water).
Even the telomere shortening studies are now looking a little dubious.
Most human studies that have observed associations between psychological stress and disease have failed to consider the very likely possibility that the proximal case of these effects is the person's health-related behaviours - loss of sleep, poor diet, drink, smoking.
PS Its best not to say 'proven'. We don't say this because you don't ever 'prove' anything in biomedical science, you just collect evidence that favours one particular hypothesis over another. Better to say, 'there is evidence to suggest'
Last edited:
Alvin
Senior Member (Voting Rights)
There might be but its beside the point and we would only waste time and effort chasing phantoms from here to the moon
BruceInOz
Senior Member (Voting Rights)
Since logit(p) is a map from the range 0->1 to -infinity->infinity, saying logit(p) is some number (any number) is a way to guarantee p is between 0 and 1 and hence interpretable as a probability. Is there anything more to it than that?
Making it a number between 0 and 1 doesn't make it a probability just a number between 0 and 1. But that is likely to be the reasoning.
In neural networks the logistic function is used to introduce a non-linearity and also limit the output to between 0 and 1 although these days I think a softmax function is often used on the output for classifiers these days.
Inara
Senior Member (Voting Rights)
Axiomatic definition of probability after Kolmogorov (note there are also other definitions, but this one was used where I studied and I guess it's the definition widely applied):
First axiom
The probability of an event is a non-negative real number:
where
is the event space. In particular,
is always finite, in contrast with more general measure theory. Theories which assign negative probability relax the first axiom.
Second axiom
This is the assumption of unit measure: that the probability that at least one of the elementary events in the entire sample space will occur is 1.
Third axiom
This is the assumption of σ-additivity:
Any countable sequence of disjoint sets (synonymous with mutually exclusive events)
satisfies
Some authors consider merely finitely additive probability spaces, in which case one just needs an algebra of sets, rather than a σ-algebra. Quasiprobability distributions in general relax the third axiom.
Taken from https://en.m.wikipedia.org/wiki/Probability_axioms
First axiom
The probability of an event is a non-negative real number:
where
Second axiom
This is the assumption of unit measure: that the probability that at least one of the elementary events in the entire sample space will occur is 1.
Third axiom
This is the assumption of σ-additivity:
Any countable sequence of disjoint sets (synonymous with mutually exclusive events)
Some authors consider merely finitely additive probability spaces, in which case one just needs an algebra of sets, rather than a σ-algebra. Quasiprobability distributions in general relax the third axiom.
Taken from https://en.m.wikipedia.org/wiki/Probability_axioms
What has been shown is that stress (including psychological) induces increased secretion of Nuclear factor kappa B. By definition this factor affects DNA.
This is the result of statistical analysis of 34 "cases" studied today. The power power of the (p) value (p=probability) to discriminate between responders and non-reponders is highly significant (P= 0.0016). This is an observation. Additional studies and tests are necesary to try explaning this observation. But the observation is what it is!
Why do you believe p is a probability because if scales between 0 and 1.
I don't really understand how your paper got published. You seem to have shown that you can train a model on data to discriminate between two classes but that is not hard. I can do that with random data that I allocate arbitrary labels. You need so demonstrate the model works with unseen data or it is completely meaningless. All you have effectively done is drawn found an arbitrary hyperplain that splits the data set you have tried to split. There is no reason why it would have any meaning especially with no underlying explanation.
Your paper and claims here seem to discredit any validity of the work by having a very sloppy methodology.
Although I see you follow similar methodology with other papers.
Last edited:
So I would come back to my question of why the output of a logistic regression would be a probability? Does it meet the basic axioms. My assumption is not but i've not looked into it. I did read about logistic regression a few years ago and it seemed to be similar to a single layered neural network with a logistic output function. (i.e. a f(sum ( input * weight ) ) where input and weight are vectors.
Woolie
Senior Member
Thanks. It would be great if you could provide a reference.
Actually I've wanted to talk about "stress" for quite a while, so now might be a good time.
I think stress, as a concept, has caused a great deal of confusion in both science and in everyday life. The concept is so vague and loose, you can apply it to virtually any sort of "bad stuff" you want, and people freely generalise from one sort of "bad stuff" to another entirely different sort of "bad stuff" without blinking. Labelling it "stress" gives it an air of certainty and a sciency-ness that blinds us to the problem.
The original idea behind stress was that it was a subjectively defined phenomenon. Rather than being something out there in the environment, it was more to do with a mismatch between environmental demands and our ability to meet them. Which means we end up with a pretty circular definition - stress is stress if the person feels stressed.
Attempts to operationalise the concept for research purposes have been fraught with difficulty. Studies involving humans tend to fall into 4 groups.
Group 1: Challenging mental tasks, such as mental arithmetic, or speeded decision-making (see this metanalysis). But that isn't what stress is, is it? No-one really believes doing the daily crossword puzzle may be injurious to one's health. What these studies show is that mental exertion has effects on your cytokine production, just as physical exertion does. Physical exertion produces much more reliable changes in cytokine production. But I don't think we're going to advise otherwise healthy people to avoid daily exercise either, are we?
Group 2: Tasks that are challenging and genuinely anxiety-provoking, at least for some people - for example, public speaking. This is a bit better, because it involves a challenge that some people may feel ill-equipped to deal with. But again, we're not really going to be advising people to avoid all situations where they might feel nervous and ill-equipped, are we? These tasks lack another component that's critical to the concept of stress - chronicity.
Group 3: "Naturalistic" studies that ask people (usually undergrads) to describe the events in, say, the last year or so of their lives. Then scores from those interviews (or questionnaires) are used to sort people into "stressed" and "not stressed". The problem is that so many things are confounded here, most especially socioeconomic status, but there's also the lifestyle confound. If you want to claim that people experiencing emotional difficulties are likely to neglect their own health needs and suffer as a result, I'm fine with that, but these studies are usually claiming something so much stronger and more direct.
Group 4: Studies examining people in life situations that are genuinely physically and/or emotionally difficult. For example, caring for a loved one with dementia. Here we're getting closer to the core idea under stress, but its still a mixed up mess. Caring for a seriously ill person is not only emotionally challenging, its also practically demanding (being on-call virtually 24/7, the carer may not adequately attend to sleep, rest and their other health needs). So its pretty hard to tease apart what's what. Perhaps this is why these studies produce such conflicting findings (some have shown increased susceptibility to colds in these people, suggesting a generally inflammatory response, but others examining response to innoculations have not found any differences).
In a word, its a mess. And that's not even considering the broad generalisations we make from animal "stress" studies (which generally involve very physical stressors, like being starved or tortured or having to swim for your life for protracted periods).
I think we need to move away from "stress" and towards more precise definitions of the psychological states we're referring to.
Woolie
Senior Member
Okay, I did worry it was a bit of a digression.
Inara
Senior Member (Voting Rights)
This needs to be checked. I never checked it, and I guess I would have to dig into logistic regression. I don't think it's trivial - a problem might be axiom 3. After seeing the definition of logit and reading a bit (and not understanding entirely!), e.g.
(https://www.quora.com/Why-is-the-output-of-logistic-regression-interpreted-as-a-probability)
I take it that logit itself is not a probability, but that the output needs to be mapped in order to receive a conditional probability. It also seems that often the output of logistic regression is interpreted as a (conditional) probability - which, it seems, could only be possible if the mapping of output to conditional probability is linear (-> axiom 3?). That's a big constraint (that's often ignored or forgotten).
I would say there are infinitely many mappings from a space to the interval [0,1]. Not every mapping will be a probability of course.
Thank you, those participant(s) who consider my scientific work and publications "shoddy". Nonetheless, my work ranks at the 97th percentile of scientific publications according to "Research Gate". Clearly, I work with patients, and ethical reasons limit the kind of "clinical experiments" I can do. In addition, pragmatic clinical research is as valuable as so-called evidence based research (see the recent publications and comments on this subject on internet).
The semantic discussion about "what is stress" is futile. It is the reaction of the body to "stress" that is important. This reaction can be measured in many ways: changes in the autonomous system with tachycardia, increased blood pressure, increased transpiration, enhanced bowel activity or contraction of the urine bladder, bronchospasm, etc. It can also be measured by endocrine markers, such as cortisol and prolactin.
Quite remarkably, there is an important difference between different persons regarding the "biological" stress-reaction to similar stressors. This is generally considered as "stress sensitivity" depending on "stress management skills". That stress can cause elevated levels of inflammatory cytokines and of nuclear factor-kappa B, inducing DNA damage has indeed been "proven" (for recent review see e.g. Wang W, Mani AM, Wu Z-H, J Cancer Metastasis Treat 2017; 3: 45-59).
Stress can deregulate both the thalamic and hypothalamic functions, influence the immune reaction, has desastrous effects on the vascular endothelium and heart, and can cause epigenetic changes in humans. It also deregulates the defence mechanism by e.g. heat shock proteins.
The semantic discussion about "what is stress" is futile. It is the reaction of the body to "stress" that is important. This reaction can be measured in many ways: changes in the autonomous system with tachycardia, increased blood pressure, increased transpiration, enhanced bowel activity or contraction of the urine bladder, bronchospasm, etc. It can also be measured by endocrine markers, such as cortisol and prolactin.
Quite remarkably, there is an important difference between different persons regarding the "biological" stress-reaction to similar stressors. This is generally considered as "stress sensitivity" depending on "stress management skills". That stress can cause elevated levels of inflammatory cytokines and of nuclear factor-kappa B, inducing DNA damage has indeed been "proven" (for recent review see e.g. Wang W, Mani AM, Wu Z-H, J Cancer Metastasis Treat 2017; 3: 45-59).
Stress can deregulate both the thalamic and hypothalamic functions, influence the immune reaction, has desastrous effects on the vascular endothelium and heart, and can cause epigenetic changes in humans. It also deregulates the defence mechanism by e.g. heat shock proteins.
Last edited: