Standard Deviation in Means Testing
Tuesday November 18, 2008 by Paul Ponsford
I am working with a research company that created a report and did all their significance testing on the means using a method I've never heard of before.
Instead of using the actual standard deviation, they used a value of 1 to determine significant differences. I'm an SPSS user and know that the program considers the actual standard deviation.
Is this a standard practice in marketing research that I am not aware of?
What are the benefits and limitations of this approach?
Thanks in advance,
Paul
» Post to this Topic
No it's not...
Tuesday November 18, 2008 by Ian L. Straus
This is a reply to: Standard Deviation in Means Testing
No it's not usual. At least unless the sample size is great enough, or the confidence level is relaxed enough, that a difference of 1? 1%? would be significant.
I think it would be appropriate for you to ask how they got to that criterion.
» Reply to this Post
I concur
Tuesday November 18, 2008 by Miles Pongrass
This is a reply to: No it's not...
I agree wtih Ian. Ask them and see what they say. Then, if possible, share their response. I would like to hear about their technique.
» Reply to this Post
Still Wondering
Wednesday November 19, 2008 by Paul Ponsford
This is a reply to: Standard Deviation in Means Testing
Thanks for the input. I have asked about why a value of 1 was used.
I am still wondering if anyone has any input as to the benefits of this approach over using the actual standard deviation.
Paul
» Reply to this Post
A Guess
Thursday December 4, 2008 by David J. Mangen
This is a reply to: Still Wondering
If I had to guess -- and it is strictly a guess -- I would suspect that it was done for the sake of simplicity and ease of understanding on the part of consumers of the research who do not grasp the concept of variance very well.
Is it legitimate? If the sample and scale are such that differences between means of 1.0 would often end up being significant, then I guess I would say that it might have some modest legitimacy. Note, however, that this is assuming that someone has pre-processed the data to come up with this approximate rule of thumb. The unfortunate part of it is that the bozo who is consuming the research will probably get it in his/her mind that a difference of 1.0 is the magic value for all research.
/* Rant on
The dumbing down of America is becoming quite pathetic.
/* Rant off
» Reply to this Post
