Normality

krd

This is a question about normal distribution (Gaussian) in statistical psychology (those questionnaires)


I may as well put the graphic here:




The question is, when you graph the answers to those questionnaires, do you always, or most often, get a bell curve?

Can you get more than one bell curve in a result?

Or can the results be wildly inconclusive - Monte Carlo?

JuliusCaesar

krd wrote: »

This is a question about normal distribution (Gaussian) in statistical psychology (those questionnaires)...
The question is, when you graph the answers to those questionnaires, do you always, or most often, get a bell curve?

What questionnaires?

krd

JuliusCaesar wrote: »

What questionnaires?

In general I mean any psychometric data - any form of sampling where you could preform statistical analysis on the results - or where that is your intention.


You know what I mean.

Strongly agree, agree, neither agree nor disagree, disagree, strongly disagree.

cranks

Quincunx. Galton. Possibly related.

https://www.youtube.com/watch?v=AUSKTk9ENzg

Mind the skew tho.


Beer, eh?

krd

cranks wrote: »

Quincunx. Galton. Possibly related.

Yes, closely related. The Galton machine gives you a bell curve for randomness. If you get 10,000 people into a football field, and measure their heights, you'll get a bell curve just like the one for Galton's machine.

And you can do it for lots of things, and you'll get the same bell curve. You can draw all kinds of inferences from seeing those curves. Like the distribution on Galtons curve looks ordered (because it produces the same curve again and again) - even though it's using randomness.

Mind the skew tho.

Skew is interesting. It can reveal some other kind of mechanism at play where a theory might suggest a system should be completely random - give a Galton like bell shaped curve.

And example would be Blackbody radiation. Studying the curve led to quantum physics. It really had them scratching their heads and pulling their hair out.

Beer, eh?

The drunkards walk.


Okay, maybe to frame my question more clearly. Say, if you took a sample group. And you put sensors all over their heads and faces. And then - Ah I may as well make this more fun - you fitted them with plethysmography devices. Then played them a few minutes of gratuitous extreme hardcore gay pornography.

Are you following me?

Then the X-access of my graph would be arousal/anxiety/revulsion etc.

What I would expect is two bumps on my graph. A little bump for the homosexuals, and a big bump for the horrified heterosexuals.

Do you get what I'm on about?

cranks

krd wrote: »

Okay, maybe to frame my question more clearly. Say, if you took a sample group. And you put sensors all over their heads and faces. And then - Ah I may as well make this more fun - you fitted them with plethysmography devices. Then played them a few minutes of gratuitous extreme hardcore gay pornography.

Are you following me?

Then the X-access of my graph would be arousal/anxiety/revulsion etc.

What I would expect is two bumps on my graph. A little bump for the homosexuals, and a big bump for the horrified heterosexuals.

Do you get what I'm on about?

Given the title you gave to the thread ('Normality') and the above hypothetical experiment you subsequently cite by way of clarification of your question, my best guess is that you're exploring the idea of what is behaviourally 'normal/abnormal' and going on to invoke statistical notions of normality.

A normal distribution, as you know, includes and allows for all persuasions.

Best I can do is to suggest the idea of kurtosis if you're talking of symmetrical distributions that deviate from typical Gaussian curves (i.e your 'little bumps and big bumps')....?

Short of that, I'm not sure I do get what you're on about, I'm afraid.


(p.s. my beer reference the other night was beer induced!:D)

krd

cranks wrote: »

Given the title you gave to the thread ('Normality') and the above hypothetical experiment you subsequently cite by way of clarification of your question, my best guess is that you're exploring the idea of what is behaviourally 'normal/abnormal' and going on to invoke statistical notions of normality.

I thought the term 'normality' would be eye catching. And that is the question I was making.

A normal distribution, as you know, includes and allows for all persuasions.

Well, what I might be expecting is you might have a majority and minority persuasion. To peaks in different places. One bell curve smaller than the other.

I think it's worrying. What does it mean, if your thinking is say a few standard deviations out from the mean?

Best I can do is to suggest the idea of kurtosis if you're talking of symmetrical distributions that deviate from typical Gaussian curves (i.e your 'little bumps and big bumps')....?

It's not really the symmetry of curve.....I would always expect the curve to be assymettric...........In any kind of test, say judging social values, you're not going to get a peak at 'neither agree nor disagree' ... And from my little reading, there's always a bias away from extremes. So, the result will always be skewed and assymmetric. If you correct for biases, as set your mean to the centre - you may still get the perfect bell curve.

Short of that, I'm not sure I do get what you're on about, I'm afraid.

Does anyone...ever.

(p.s. my beer reference the other night was beer induced!:D)

Yes, wonderful stuff isn't it.