Welcome, Guest. Please login or register.

July 17, 2019, 10:23:45 am

Author Topic: P value  (Read 1190 times)  Share 

0 Members and 1 Guest are viewing this topic.

#J.Procrastinator

  • Victorian
  • Trendsetter
  • **
  • Posts: 165
  • Respect: 0
  • School Grad Year: 2014
P value
« on: October 20, 2013, 11:18:09 pm »
0
What exactly are P values and in what ways can we talk about them?

I'm assuming that we need to know them for the extended response section C part. Does it have to do with the generalisation of results also?

Thank you!
2015-2017: Bachelor of Science @ UoM

TrueTears

  • TT
  • Honorary Moderator
  • It's Over 9000!!
  • *******
  • Posts: 16358
  • Respect: +653
Re: P value
« Reply #1 on: October 20, 2013, 11:30:46 pm »
+2
I like the following discussion of p value on CV: http://stats.stackexchange.com/questions/44769/understanding-p-value?rq=1. The post by Zag is especially informative.

For examples and illustrations, I direct you to http://stats.stackexchange.com/questions/44999/p-value-with-multimodal-pdf-of-a-test-statistic
« Last Edit: October 20, 2013, 11:54:24 pm by TrueTears »

alondouek

  • Subject Review God
  • Honorary Moderator
  • ATAR Notes Superstar
  • *******
  • Posts: 2912
  • Oh to be a Gooner!
  • Respect: +312
  • School: Leibler Yavneh College
  • School Grad Year: 2012
Re: P value
« Reply #2 on: October 20, 2013, 11:38:44 pm »
+1
I don't know if p-values are in the psych course (having never done psych) - but if you like, here is some background info:

Spoiler
p-values are sorta-abstract concepts.

Basically, the p-value (p for probability) is the likelihood of obtaining a value as or more extreme than the data you have, assuming that the null hypothesis () for whatever you're investigating is true (usually that nothing has changed between two variables, or that there is no association between two variables). The opposite of the null hypothesis is the "alternative hypothesis", or .


To use p-values, we need a significance level (usually designated ). In statistics, a common significance level is 95%, or .

If we use this 95% confidence/significance level, then we can say the following:

  • If p-value is less than 0.05, reject in favour of (often suggesting there is a significant relationship between two variables etc)
  • If p-value is greater than 0.05, DO NOT reject (suggesting that there is no association between two variables etc)

We obtain p-values from collected/observed data by performing statistical tests called hypothesis tests and confidence intervals. I don't know if you cover these in psych, but they're very useful. N.B. there are many different types of hypothesis test based on the types of variables (categorical/numerical) and the situation the data is collected from.

If you read but didn't understand the spoiler-tag stuff, I'd be happy to elaborate as much as I can :)
« Last Edit: October 20, 2013, 11:40:59 pm by alondouek »
2013-2016
Majoring in Genetics and Developmental Biology

2012 ATAR: 96.55
English [48] Biology [40]

Need a driving instructor? Mobility Driving School

#J.Procrastinator

  • Victorian
  • Trendsetter
  • **
  • Posts: 165
  • Respect: 0
  • School Grad Year: 2014
Re: P value
« Reply #3 on: October 21, 2013, 12:31:34 am »
0
Thanks TrueTears for the helpful links, and alondouek for the informative explanation! I've got a general idea of what they are now :) These seem more maths related? What math subjects do P values come up in?

In psychology, it does relate to these concepts, but only briefly. I think it's more so to do with the chances of a particular situation occurring and whether the results can be generalised to the wider population.
« Last Edit: October 21, 2013, 12:33:35 am by #J.Procrastinator »
2015-2017: Bachelor of Science @ UoM

alondouek

  • Subject Review God
  • Honorary Moderator
  • ATAR Notes Superstar
  • *******
  • Posts: 2912
  • Oh to be a Gooner!
  • Respect: +312
  • School: Leibler Yavneh College
  • School Grad Year: 2012
Re: P value
« Reply #4 on: October 21, 2013, 01:05:33 am »
0
Thanks TrueTears for the helpful links, and alondouek for the informative explanation! I've got a general idea of what they are now :) These seem more maths related? What math subjects do P values come up in?

In psychology, it does relate to these concepts, but only briefly. I think it's more so to do with the chances of a particular situation occurring and whether the results can be generalised to the wider population.

You're welcome! p-values are found in statistics; simply put (for the purposes of VCE psych), they're used to determine if a given claim/statement (a hypothesis) is likely to be correct.

You'll have learned in psych (I think) about taking sample measurements, and then using statistics to infer characteristics about the population from which we derive those samples, based on the samples themselves. p-values (as a part of hypothesis testing) are used to test the validity of our generalising sample statistics (which we get from the data we collect) to population parameters (i.e. what is really happening on a large, population scale).

P-values and hypothesis testing have a huge foundation of very complicated mathematics (which TT could tell you more about, it's one of his areas of specialty), but you don't need to know much of this foundation to apply these statistical techniques and understand the results you get from them.
2013-2016
Majoring in Genetics and Developmental Biology

2012 ATAR: 96.55
English [48] Biology [40]

Need a driving instructor? Mobility Driving School

#J.Procrastinator

  • Victorian
  • Trendsetter
  • **
  • Posts: 165
  • Respect: 0
  • School Grad Year: 2014
Re: P value
« Reply #5 on: October 21, 2013, 04:05:30 pm »
0
You're welcome! p-values are found in statistics; simply put (for the purposes of VCE psych), they're used to determine if a given claim/statement (a hypothesis) is likely to be correct.

You'll have learned in psych (I think) about taking sample measurements, and then using statistics to infer characteristics about the population from which we derive those samples, based on the samples themselves. p-values (as a part of hypothesis testing) are used to test the validity of our generalising sample statistics (which we get from the data we collect) to population parameters (i.e. what is really happening on a large, population scale).

P-values and hypothesis testing have a huge foundation of very complicated mathematics (which TT could tell you more about, it's one of his areas of specialty), but you don't need to know much of this foundation to apply these statistical techniques and understand the results you get from them.

Thank you very much :)
2015-2017: Bachelor of Science @ UoM

brenden

  • Honorary Moderator
  • Great Wonder of ATAR Notes
  • *******
  • Posts: 7150
  • Respect: +2511
Re: P value
« Reply #6 on: October 21, 2013, 07:05:16 pm »
0
To give you a more Psych related answer, you will be expected to know that a p-value of <.05 means that it is statistically significant. What <.05 means is that less than 5 experiments out of 100 would have occured to chance (so 95% of the time, it was the independent variable that correlated with the results and the results didn't happen by chance).

Watch out for multi-choice answers that say ">.05, <.5". That says greater than 5 in 100 trials, and less than 50 of 100 trials, respectively.

In a short answer question, "what does a p value of <.0x mean?" You just need to write "This p-value means that in this experiment x amount of times the results could have occurred due to chance". (Usually it will be <0.05, but they will sometimes change the number, which is why I used 'x'... Moreover, some studies will say "for this the be statistically significant, we want a p-value of <0.01 (1 trail due to chance out of 100).
This means that the lower value of the p-value, the better.


*Just skimmed alon's and TT's, sorry for any repetition (or any incorrect information; I wouldn't put it past VCAA to make the requirements of the course something different to how it is in reality).
✌️just do what makes you happy ✌️

slothpomba

  • Honorary Moderator
  • ATAR Notes Legend
  • *******
  • Posts: 4462
  • Chief Executive Sloth
  • Respect: +326
Re: P value
« Reply #7 on: October 21, 2013, 07:30:55 pm »
0
I've had a few epidemiology lectures so i think i'm barely qualified to chime in (still wont stop me!).

Basically the P values tells you the probability that your result is just by sheer chance and not a cause/effect thing in terms of what you're looking at.

Let's say i get a bunch of people to do a rain dance and look at the results. It rains 3 days out of the 7 they do the rain dance. What is the probability the raining is caused by their little dance and simply not just by it happening to rain anyway? Let's say we get a P value of 0.90 . In terms of science, thats pretty abysmally high. It usually rains 2.8 days on average in a week (just made that up). It rained 3 days a week when we did our intervention. That's not much more than you'd expect by chance, so, its 90% likely that this result was just down to chance.

Let's take another example (a real one). Mirtazapine reduces depression significantly more than other drugs at 2 weeks. It could just be down to variability/chance that people taking mirtazapine got better faster. How do we decide whether it is or isn't? We look at the P value, in this case it was P < 0.00001. This means there is less than a 0.001% chance that this effect is simply due to chance, that is absolutely tiny, we can be almost (but not entirely!, this delves more into philosophy/stats though) certain that this drug does indeed work faster.

I'm sure you can imagine it in terms of all kinds of things. Say the P value that smoking causes lung cancer, would you expect this to be high or low? The P value of wearing red and being late to work, high or low? What degree does A cause B here compared to how much its just down to chance?
« Last Edit: October 21, 2013, 07:32:30 pm by slothpomba »

ATAR Notes Chat
Philosophy thread
-----
2011-15: Bachelor of Science/Arts (Religious studies) @ Monash Clayton - Majors: Pharmacology, Physiology, Developmental Biology
2016: Bachelor of Science (Honours) - Psychiatry research

#J.Procrastinator

  • Victorian
  • Trendsetter
  • **
  • Posts: 165
  • Respect: 0
  • School Grad Year: 2014
Re: P value
« Reply #8 on: October 21, 2013, 07:48:30 pm »
0
To give you a more Psych related answer, you will be expected to know that a p-value of <.05 means that it is statistically significant. What <.05 means is that less than 5 experiments out of 100 would have occured to chance (so 95% of the time, it was the independent variable that correlated with the results and the results didn't happen by chance).

Watch out for multi-choice answers that say ">.05, <.5". That says greater than 5 in 100 trials, and less than 50 of 100 trials, respectively.

In a short answer question, "what does a p value of <.0x mean?" You just need to write "This p-value means that in this experiment x amount of times the results could have occurred due to chance". (Usually it will be <0.05, but they will sometimes change the number, which is why I used 'x'... Moreover, some studies will say "for this the be statistically significant, we want a p-value of <0.01 (1 trail due to chance out of 100).
This means that the lower value of the p-value, the better.


*Just skimmed alon's and TT's, sorry for any repetition (or any incorrect information; I wouldn't put it past VCAA to make the requirements of the course something different to how it is in reality).

I've had a few epidemiology lectures so i think i'm barely qualified to chime in (still wont stop me!).

Basically the P values tells you the probability that your result is just by sheer chance and not a cause/effect thing in terms of what you're looking at.

Let's say i get a bunch of people to do a rain dance and look at the results. It rains 3 days out of the 7 they do the rain dance. What is the probability the raining is caused by their little dance and simply not just by it happening to rain anyway? Let's say we get a P value of 0.90 . In terms of science, thats pretty abysmally high. It usually rains 2.8 days on average in a week (just made that up). It rained 3 days a week when we did our intervention. That's not much more than you'd expect by chance, so, its 90% likely that this result was just down to chance.

Let's take another example (a real one). Mirtazapine reduces depression significantly more than other drugs at 2 weeks. It could just be down to variability/chance that people taking mirtazapine got better faster. How do we decide whether it is or isn't? We look at the P value, in this case it was P < 0.00001. This means there is less than a 0.001% chance that this effect is simply due to chance, that is absolutely tiny, we can be almost (but not entirely!, this delves more into philosophy/stats though) certain that this drug does indeed work faster.

I'm sure you can imagine it in terms of all kinds of things. Say the P value that smoking causes lung cancer, would you expect this to be high or low? The P value of wearing red and being late to work, high or low? What degree does A cause B here compared to how much its just down to chance?

Thanks both for your detailed input! I understand it better now  :D
2015-2017: Bachelor of Science @ UoM