type I error and p-value

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

type I error and p-value

Zdaniuk, Bozena-3
hello, can anyone offer a straightforward explanation (or point me to one) on why type I error and p-value are NOT the same thing?
thanks so much,
bozena
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Reply | Threaded
Open this post in threaded view
|

Re: type I error and p-value

Mario Giesel-2
How about:

type I error = Rejecting a true null hypothesis
p = probability of rejecting a true null hypothesis

Mario Giesel
Munich, Germany


Am Dienstag, 7. Mai 2019, 07:32:13 MESZ hat Zdaniuk, Bozena <[hidden email]> Folgendes geschrieben:


hello, can anyone offer a straightforward explanation (or point me to one) on why type I error and p-value are NOT the same thing?
thanks so much,
bozena
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Reply | Threaded
Open this post in threaded view
|

FW: type I error and p-value

Zdaniuk, Bozena-3

Thank you, Mario.

So, shouldn’t the p-value be considered the probability of Type I error?

b

 

From: Mario Giesel <[hidden email]>
Sent: May 7, 2019 2:04 AM
To: [hidden email]; Zdaniuk, Bozena <[hidden email]>
Subject: Re: type I error and p-value

 

How about:

 

type I error = Rejecting a true null hypothesis

p = probability of rejecting a true null hypothesis

 

Mario Giesel

Munich, Germany

 

 

Am Dienstag, 7. Mai 2019, 07:32:13 MESZ hat Zdaniuk, Bozena <[hidden email]> Folgendes geschrieben:

 

 

hello, can anyone offer a straightforward explanation (or point me to one) on why type I error and p-value are NOT the same thing?

thanks so much,

bozena

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Reply | Threaded
Open this post in threaded view
|

Re: FW: type I error and p-value

Bruce Weaver
Administrator
If p > alpha, you would not reject H0 if you follow the usual practice.  So
that definition runs into some difficulty.  I think the following definition
of p is better.  

p = the probability of getting a test statistic at least as extreme as the
observed value if the null hypothesis is true.

Notice that this is a conditional probability--conditional on H0 being true.

Finally, this recent special issue of The American Statistician may be of
interest.

  https://www.tandfonline.com/toc/utas20/73/sup1?nav=tocList

Here is one interesting (and provocative) recommendation in the editorial
that leads off the special issue.

--- start of excerpt ---
The ASA Statement on P-Values and Statistical Significance stopped just
short of recommending that declarations of "statistical significance" be
abandoned. We take that step here.  We conclude, based on our review of the
articles in this special issue and the broader literature, that it is time
to stop using
the term "statistically significant" entirely. Nor should variants such as
"significantly different," "p < 0.05," and "nonsignificant" survive, whether
expressed in words, by asterisks in a table, or in some other way.
--- end of excerpt ---

HTH.


Zdaniuk, Bozena-3 wrote
> Thank you, Mario.
> So, shouldn’t the p-value be considered the probability of Type I error?
> b
>
> From: Mario Giesel &lt;

> spss.giesel@

> &lt;mailto:

> spss.giesel@

> &gt;>
> Sent: May 7, 2019 2:04 AM
> To:

> SPSSX-L@.UGA

> &lt;mailto:

> SPSSX-L@.UGA

> &gt;; Zdaniuk, Bozena &lt;

> bozena.zdaniuk@

> &lt;mailto:

> bozena.zdaniuk@

> &gt;>
> Subject: Re: type I error and p-value
>
> How about:
>
> type I error = Rejecting a true null hypothesis
> p = probability of rejecting a true null hypothesis
>
> Mario Giesel
> Munich, Germany
>
>
> Am Dienstag, 7. Mai 2019, 07:32:13 MESZ hat Zdaniuk, Bozena &lt;

> bozena.zdaniuk@

> &lt;mailto:

> bozena.zdaniuk@

> &gt;> Folgendes geschrieben:
>
>
> hello, can anyone offer a straightforward explanation (or point me to one)
> on why type I error and p-value are NOT the same thing?
> thanks so much,
> bozena
> ===================== To manage your subscription to SPSSX-L, send a
> message to

> LISTSERV@.UGA

> &lt;mailto:

> LISTSERV@.UGA

> &gt; (not to SPSSX-L), with no body text except the command. To leave the
> list, send the command SIGNOFF SPSSX-L For a list of commands to manage
> subscriptions, send the command INFO REFCARD
>
> =====================
> To manage your subscription to SPSSX-L, send a message to

> LISTSERV@.UGA

>  (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD





-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.
Reply | Threaded
Open this post in threaded view
|

Re: FW: type I error and p-value

Kirill Orlov
Bozena,
p-value is the expected, a priori conditional (H0 is true) type I error probability.
I is not that aposteriori type I error probability when you've conducted the experiment.
 


07.05.2019 22:51, Bruce Weaver пишет:
If p > alpha, you would not reject H0 if you follow the usual practice.  So
that definition runs into some difficulty.  I think the following definition
of p is better.  

p = the probability of getting a test statistic at least as extreme as the
observed value if the null hypothesis is true.

Notice that this is a conditional probability--conditional on H0 being true.

Finally, this recent special issue of The American Statistician may be of
interest.

  https://www.tandfonline.com/toc/utas20/73/sup1?nav=tocList

Here is one interesting (and provocative) recommendation in the editorial
that leads off the special issue.

--- start of excerpt ---
The ASA Statement on P-Values and Statistical Significance stopped just
short of recommending that declarations of "statistical significance" be
abandoned. We take that step here.  We conclude, based on our review of the
articles in this special issue and the broader literature, that it is time
to stop using
the term "statistically significant" entirely. Nor should variants such as
"significantly different," "p < 0.05," and "nonsignificant" survive, whether
expressed in words, by asterisks in a table, or in some other way.
--- end of excerpt ---

HTH.


Zdaniuk, Bozena-3 wrote
Thank you, Mario.
So, shouldn’t the p-value be considered the probability of Type I error?
b

From: Mario Giesel &lt;
spss.giesel@
&lt;mailto:
spss.giesel@
&gt;>
Sent: May 7, 2019 2:04 AM
To: 
[hidden email]
&lt;mailto:
[hidden email]
&gt;; Zdaniuk, Bozena &lt;
bozena.zdaniuk@
&lt;mailto:
bozena.zdaniuk@
&gt;>
Subject: Re: type I error and p-value

How about:

type I error = Rejecting a true null hypothesis
p = probability of rejecting a true null hypothesis

Mario Giesel
Munich, Germany


Am Dienstag, 7. Mai 2019, 07:32:13 MESZ hat Zdaniuk, Bozena &lt;
bozena.zdaniuk@
&lt;mailto:
bozena.zdaniuk@
&gt;> Folgendes geschrieben:


hello, can anyone offer a straightforward explanation (or point me to one)
on why type I error and p-value are NOT the same thing?
thanks so much,
bozena
===================== To manage your subscription to SPSSX-L, send a
message to 
[hidden email]
&lt;mailto:
[hidden email]
&gt; (not to SPSSX-L), with no body text except the command. To leave the
list, send the command SIGNOFF SPSSX-L For a list of commands to manage
subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email]
 (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD



-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD



===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Reply | Threaded
Open this post in threaded view
|

Re: FW: type I error and p-value

Anthony Babinec

There are lots of misunderstandings and misinterpretations

around p-values and significance levels. For more on this,

see either edition of Rex Kline’s “Beyond Significance Testing.”

 

There is also a good article in the February 2019 Issue of The

American Statistician by Todd Kuffner and Stephen Walker

entitled “Why are p-values Controversial?”

 

Tony Babinec

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD