Jump to content
House Price Crash Forum

Archived

This topic is now archived and is closed to further replies.

Smell the Fear

Statistics Discussion

Recommended Posts

Hi all

I'm hoping some of the contributors to my last thread (which has been deleted for some reason rather than moved off-topic) can come here to finish the discussion we were having. It was extremely useful, particularly the contributions of Datamonkey, Daedalus, wad and Crash2006 if I recall correctly (apologies to anyone I missed).

So, the crux of what I need to know is the following:

If I have a t-test p value of e.g. 0.45 can I say that there is a 55% probability that there is a difference between the numbers I am testing (in the case of our previous discussion) and can I then say fairly that on the balance of probabilities the observed difference between two means is more likely to be genuine rather than a result of chance?

Share this post


Link to post
Share on other sites
Hi all

I'm hoping some of the contributors to my last thread (which has been deleted for some reason rather than moved off-topic) can come here to finish the discussion we were having. It was extremely useful, particularly the contributions of Datamonkey, Daedalus, wad and Crash2006 if I recall correctly (apologies to anyone I missed).

So, the crux of what I need to know is the following:

If I have a t-test p value of e.g. 0.45 can I say that there is a 55% probability that there is a difference between the numbers I am testing (in the case of our previous discussion) and can I then say fairly that on the balance of probabilities the observed difference between two means is more likely to be genuine rather than a result of chance?

No

Share this post


Link to post
Share on other sites

p-value -- A statistics term. A measure of probability that a difference between groups during an experiment happened by chance. For example, a p-value of .01 (p = .01) means there is a 1 in 100 chance the result occurred by chance. The lower the p-value, the more likely it is that the difference between groups was caused by treatment.

Share this post


Link to post
Share on other sites
p-value -- A statistics term. A measure of probability that a difference between groups during an experiment happened by chance. For example, a p-value of .01 (p = .01) means there is a 1 in 100 chance the result occurred by chance. The lower the p-value, the more likely it is that the difference between groups was caused by treatment.

This seems to confirm my reasoning that a p value of say 0.40 indicates a 40% probability that the result is due to chance, and hence a 60% probability that it is NOT due to chance?

So far most posters have seemed resistant to this conclusion but I can't quite figure out why..........

Share this post


Link to post
Share on other sites

.

z= (1.65 , -1.65) p valve= 3.5% 1.65 cuts off at 3.5%

lets say the significance level is 3.5% then we reject null can say this 96.5%

confidence.

so if it cuts of at 40% you can say 60% confidence that we have accepted.

Share this post


Link to post
Share on other sites
.

z= (1.65 , -1.65) p valve= 3.5% 1.65 cuts off at 3.5%

lets say the significance level is 3.5% then we reject null can say this 96.5%

confidence.

so if it cuts of at 40% you can say 60% confidence that we have accepted.

I think you are agreeing with me, but I'm not certain - can you clarify?

Share this post


Link to post
Share on other sites

z=1.65 the associated p value obtained from the table is 3.5%, so 1.65 cut of point is at 3.5% the null hyp can be rejected at 3.5 significant level or alternatively express with 96.5 confidence.

for a two tail test we 2 * 3.5% = 7% at this level we reject it with a confidence level of 93%.

Share this post


Link to post
Share on other sites
z=1.65 the associated p value obtained from the table is 3.5%, so 1.65 cut of point is at 3.5% the null hyp can be rejected at 3.5 significant level or alternatively express with 96.5 confidence.

for a two tail test we 2 * 3.5% = 7% at this level we reject it with a confidence level of 93%.

think I get it. When do you use 2 tail rather than 1 tail? and what does "tail" refer to exactly

for example if the data is test scores which can range from 0 to 100 and I have two groups whose means I want to compare using a t-test, should I use 1 or 2 tail?

Share this post


Link to post
Share on other sites
if alpha id .05 or 5% if you reject it or accept it then your 95% confident of the result.

is alpha the same as p?

if alpha is 0.40 can I be 60% confident of the result? I am interested in what happens when we move away from 95%

Share this post


Link to post
Share on other sites
think I get it. When do you use 2 tail rather than 1 tail? and what does "tail" refer to exactly

for example if the data is test scores which can range from 0 to 100 and I have two groups whose means I want to compare using a t-test, should I use 1 or 2 tail?

2 tail if h0 is equal to 100

h1 is not equal to 100

1 tail h0 is equal or greater than 100

h1 is less than 100

so we have one wave( bell curve) one tail would be one end of the tail.

two tail one wave( bell curve) will be both ends of the tail.

let say critical value is 1.56

we have a bell curve, the peak of the bell curve is the mean 0 the left side of the mean 0 is negative the right side is positive. two tail test your testing both side, one tail on side.

http://www.mathsrevision.net/alevel/pages.php?page=64

http://www.cliffsnotes.com/WileyCDA/Cliffs...leId-25929.html

Share this post


Link to post
Share on other sites
2 tail if h0 is equal to 100

h1 is not equal to 100

1 tail h0 is equal or greater than 100

h1 is less than 100

so we have one wave( bell curve) one tail would be one end of the tail.

two tail one wave( bell curve) will be both ends of the tail.

let say critical value is 1.56

we have a bell curve, the peak of the bell curve is the mean 0 the left side of the mean 0 is negative the right side is positive. two tail test your testing both side, one tail on side.

so if my h0 (i.e. null hypothesis?) is that the difference between two means (e.g. 60 and 61) is zero what would I use, 1 tail or 2 tail? It doesn't seem to fit your criteria. What is the significance of 100 to all this?

Thanks for your help so far.......... ;)

Share this post


Link to post
Share on other sites
so if my h0 (i.e. null hypothesis?) is that the difference between two means (e.g. 60 and 61) is zero what would I use, 1 tail or 2 tail? It doesn't seem to fit your criteria. What is the significance of 100 to all this?

Thanks for your help so far.......... ;)

100 was just a number i used, just to show you thats all , look above i found 2 links that tell you why its used, and how its used.

one tail when you have a direction < or > thats one tail, ie less than 100 or greater than 100.

2 tail = or not equal no direction if you want to find out if its equal to 100 or not equal to 100.

Share this post


Link to post
Share on other sites
so if my h0 (i.e. null hypothesis?) is that the difference between two means (e.g. 60 and 61) is zero what would I use, 1 tail or 2 tail? It doesn't seem to fit your criteria. What is the significance of 100 to all this?

Thanks for your help so far.......... ;)

1 tail is the mean less than 60 or greater than 60

2 tail is the mean = to 60 or not equal to 60

Share this post


Link to post
Share on other sites
is alpha the same as p?

if alpha is 0.40 can I be 60% confident of the result? I am interested in what happens when we move away from 95%

no

α = .05 or 5% the significant level.

so its like this

h0=100

h1=> 100

α .05 or 5%

look up critical value base on your α on stats tables

calculate test stats

decision based on the test.

Share this post


Link to post
Share on other sites
1 tail is the mean less than 60 or greater than 60

2 tail is the mean = to 60 or not equal to 60

starting to get it, I'll come back to it tomorrow. Thanks for the links, they are very good and validate my thoughts to some extent. I think I might have problems convincing most people with a rudimentary understanding of statistics as they mostly seem convinced that anything with a p<0.05 is fact and anything with a p>0.05 is fiction.

A bizarre view resulting from the fact that journals use the 0.05 cut off, nothing else! The p value properly interpreted gives a lot of info.

For example, if a new drug to cure cancer had a p of 0.30 should it be used? It has a 70% chance of having a genuine effect so it is worth trying (this needs to be balanced against side effects and cost of course). To write it off due to its p value would be foolish.

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

  • 399 Brexit, House prices and Summer 2020

    1. 1. Including the effects Brexit, where do you think average UK house prices will be relative to now in June 2020?


      • down 5% +
      • down 2.5%
      • Even
      • up 2.5%
      • up 5%



×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.