"trends" and others

Thread View: sci.math.stat

2 messages

[T] Back to Threads [V] Tree View [G] Group Articles

2 total messages Started by Shan Duncan Tue, 21 Jul 1992 16:48

#59

Author: Shan Duncan
Date: Tue, 21 Jul 1992 16:48

25 lines
1404 bytes

I would like to know how people feel about using such words as "trend".


For example:

When I see the statement:

   The data demonstrates a trend towards [fill in the blank], but this
   relationship was not significant.

Technically this is correct.   In other words I have been told by someone
that you can state a direction (i.e. less, more, greater whatever) as a trend
as long as you state if it is significant or not.  Yet at times I read this
as being "hopeful".  I have seen this type of statement used when  the person
writing the article wants to have the trend in that direction, but they did
not achive statistical significance.  

I have also seen statements such as:  There was a greater response to
[again fill in the blank] but this was not significant.

I am more comfortable with the former rather than the latter.


So what are the rules of thumb in cases such as these?

Re: "trends" and others

#63

Author: Travis Gee
Date: Wed, 22 Jul 1992 22:24

43 lines
2506 bytes

In <1992Jul21.164831.6412@bronze.ucs.indiana.edu> duncan@loris.cisab.indiana.edu (Shan Duncan) writes:
>I would like to know how people feel about using such words as "trend".
>For example:

>When I see the statement:
>   The data demonstrates a trend towards [fill in the blank], but this
>   relationship was not significant.
>Technically this is correct.   In other words I have been told by someone
>that you can state a direction (i.e. less, more, greater whatever) as a trend
>as long as you state if it is significant or not.  Yet at times I read this
>as being "hopeful". 

Yes, it is being hopeful, and those hopes should be augmented by a
proper power analysis. If the trend is there, but they only had a 15%
chance of detecting an effect of  size [f.i.t.b.] given their N and alpha,
then that should be a red flag indicating that replication is needed
with a more adequate sample. 

> I have seen this type of statement used when  the person
>writing the article wants to have the trend in that direction, but they did
>not achive statistical significance.  

Statistical signficance is a mode of rhetoric which we adopt, and which
reflects our willingness to risk Type I error. If the poor sap set his
alpha at .001, and got .004, then I'd gladly undertake a replication which
set more realistic risk-limits on the basis of his  'trend'. But if he
set it at .10, with 99% power, I'd tell him to jump in the lake if he
asked for funding.

>I have also seen statements such as:  There was a greater response to
>[again fill in the blank] but this was not significant.

>I am more comfortable with the former rather than the latter.

Me too..."trend" is sufficiently vague without being provocative. 

>So what are the rules of thumb in cases such as these?

 Stay with an alpha  that is  commonly accepted  in the  literature of
[f.i.t.b.]   and  accompany  statements of trend with a calculation of
1-Beta, so the reader can better judge what's happening. If somebody
talks about a trend without it, calculate it if you can before passing
too much judgement.

Thread Navigation

This is a paginated view of messages in the thread with full content displayed inline.

Messages are displayed in chronological order, with the original post highlighted in green.

Use pagination controls to navigate through all messages in large threads.

Back to All Threads