P value

[katykins]

i get the general gist of them, but could someone explain to me about how they are statistically significant ect. what does the p value have to be to support the hypothesis? i feel like all the research methods material has slipped through the cracks for me

[brenden]

i get the general gist of them, but could someone explain to me about how they are statistically significant ect. what does the p value have to be to support the hypothesis? i feel like all the research methods material has slipped through the cracks for me

Okay so a P value of <0.05 is 'less than 5 in 100 trials could occur to chance' -> if it is equal to, or less than the designated P value (they change them rarely) then it means the study is statistically significant and can be generalised back to the -research- population.

The decimal is used to refer to 100. So if we take '1' as 100%, then .5 must be 50%, thus .05 must be 5%. So 5% of trials could occur do to chance
Essentially, what a significant p value is saying is: the independent variable was the influence on the dependent variable, and the variation in the dependent variable -wasn't- due to chance.

[TrueTears]

The p value is the probability of obtaining a test statistic at least as extreme as the one that was actually observed, assuming that the null hypothesis is true.

Note that if p value < alpha then we reject the null hypothesis and if p value > alpha then we do not reject the null hypothesis (note that we never ever say we ACCEPT the null hypothesis).

alpha is the significance level, by convention we set it to be 5% or 1%, and you can use different mathematical techniques, such as the maximum likelihood method or the Blackwell-Rao theorem (http://en.wikipedia.org/wiki/Rao%E2%80%93Blackwell_theorem) to generate the most "powerful" test.

Note that we can also say that if p value < alpha then it is statistically significant, if not, then it is statistically insignificant. The p value derives from the distribution of the test statistic, eg, t distribution, F distribution, chi squared and so on

Note that for statistical inferences we can never be sure whether the inference is correct or not (for the population), we often characterise this by examining errors, Type I or Type II, http://en.wikipedia.org/wiki/Type_I_and_type_II_errors#Type_I_error, thus there are often great tools to maximise the probability of NOT obtaining a certain type of error or minimising the chances of getting a type of error.

[abeybaby]

the p value is the probability that you observe a difference between conditions GIVEN THAT the null hypothesis (null=nothing, so the null hypothesis is the hypothesis that whatever you changed, did nothing) is true. e.g.,

If you take 1000 people, and give them panadol, and measure their sensitivity to pain,
and another 1000, give them a placebo and measure their sensitivity, we EXPECT a difference.

say we do find a difference.

The p value gives you the probability of seeing that particular difference (or more) IFFFFF there was in fact, no difference between panadol and the placebo. so if you get a really high p value, that means that the null hypothesis was probably true, and panadol doesnt really do anything

if p was TINY then there was SUCH LITTLE chance of seeing such a big difference, because were assuming that the null hypothesis is true (ie, that panadol is useless). therefore, your original assumption, that panadol sucks, is probably wrong, and we have a "significant" effect.

as a disciplne, psychology is prepared to accept p values of less than 0.05. so that means, that theyre prepared to be wrong 1 in every 20 times.

hope this helps!

EDIT: TT beat me t oit, and has a better explanation!

[Genericname2365]

I had a post as well but I think the above posts should explain it between them haha. 

EDIT: TT beat me t oit, and has a better explanation!

Happened to me three times in a row in this thread....lol

[totaled]

These are my notes from 2-3 years back on p values hopefully they'll help if you haven't understood it yet from all the responses above:


P value is a probability value of the significance level of any difference.
-> Generally accepted for 0.05. If the p value ≥ 0.05, it means the difference in mean scores obtained was most likely due to chance and this would be unacceptable. It would therefore be said that the results did not support the hypothesis and a conclusion would be made that this difference obtained was most likely due to chance alone.
Example: p = 0.01, is a low value, indicating a probability of 1% that results are due to chance alone.

Statistical significance is the term used to indicate whether the difference in the results obtained for the control and experimental groups is a real difference, due to the independent variable, and probably not due to chance factors alone. In any experiment, the mean scores from each group will not be identical. If the likelihood of the difference occurring by chance is extremely low, then it is said to be statistically significant (if the p value is low, the results are considered as ‘statistically significant’)

[sheep94]

About the p-value, if you're writing a report on an experiment and you're given the p-value of <0.05 but there are extraneous variables e.g. convenience sampling, should you still accept the hypothesis?

[RTandon]

About the p-value, if you're writing a report on an experiment and you're given the p-value of <0.05 but there are extraneous variables e.g. convenience sampling, should you still accept the hypothesis?

Nope (:

[TrueTears]

About the p-value, if you're writing a report on an experiment and you're given the p-value of <0.05 but there are extraneous variables e.g. convenience sampling, should you still accept the hypothesis?

There is no such thing as "Accepting" the (null) hypothesis, you either reject it or cannot reject it, and yes - there is a big difference between saying accepting the null and cannot rejecting it, the reason is more mathematical based however the wording here is not trivial, it is rather important.

[pi]

These are my notes from 2-3 years back on p values hopefully they'll help if you haven't understood it yet from all the responses above

You should have notes from those dreaded epi tutes as well