Skeptic Friends Network

Username:
Password:
Save Password
Forgot your Password?
Home | Forums | Active Topics | Active Polls | Register | FAQ | Contact Us  
  Connect: Chat | SFN Messenger | Buddy List | Members
Personalize: Profile | My Page | Forum Bookmarks  
 All Forums
 Our Skeptic Forums
 Pseudoscience
 About that 655k dead Iraqi number...
 New Topic  Topic Locked
 Printer Friendly Bookmark this Topic BookMark Topic
Previous Page
Author Previous Topic Topic Next Topic
Page: of 4

chaloobi
SFN Regular

1620 Posts

Posted - 11/03/2006 :  12:35:59   [Permalink]  Show Profile  Send chaloobi a Yahoo! Message Send chaloobi a Private Message
Back to statistics and testing.....

From my own experience, you can tell a lot about a data set if you plot the distribution. If it's normal (bell shaped) then you've likely got a good sample. However, if it's skewed one way or another or if it's all over the place, then there is some problem or bias in the system you're studying. That or your Measurement System is faulty. And that's really what we're talking about here. Is the measurement methodology used in this study impartial?

Typically what's done is a Measurement System Analysis to verify the data collection is impartial, non-biased, etc. That's done before any sample data is even taken. Back end verification that your data is unbiased is first to look at the distribution of the data points. If it's normal (bell shaped) with all center measures equal (mean, median and mode) then you've got a good sample. However, if it's skewed one way or another or if it's all over the place, then you can say there was some problem or bias in the system you're studying (which can be accounted for). That or your measurement system is faulty. Probably there's any number of ways to plot this data and test for normality - by location, by type of death, by whatever....

I did some looking for information on statistical tests for validity of data and came up with this web site which is a huge statistics resource. It's got tons of stuff, so check it out if you're interested in stats.

http://home.ubalt.edu/ntsbarsh/stat-data/Topics.htm

I browsed through the topics to find some info on what kinds of tests that might apply to data validity, etc. I came up with:

A. Tests for Normality, RE what I was talking about above:
http://home.ubalt.edu/ntsbarsh/stat-data/Topics.htm#rTestNormal

B. This is a technique for determining the reliablity of your data gathering methods (raters in this case would be the groups of doctors doing surveys). They compare data gathered between the 'raters' and even between different sets taken by the same rater, looking for bias.
http://home.ubalt.edu/ntsbarsh/stat-data/Topics.htm#riirr

C. This is apparently a widely used test of data:
http://home.ubalt.edu/ntsbarsh/stat-data/Topics.htm#rCohenKappa

D. Bias reduction techniques:
http://home.ubalt.edu/ntsbarsh/stat-data/Topics.htm

These articles are brief and geared for stuedents of statistics (ie not for the layman). But they give you an idea of what statistics can do and at a minimum they introduce ideas and terms you can use to do further searches.

-Chaloobi

Go to Top of Page
Page: of 4 Previous Topic Topic Next Topic  
Previous Page
 New Topic  Topic Locked
 Printer Friendly Bookmark this Topic BookMark Topic
Jump To:

The mission of the Skeptic Friends Network is to promote skepticism, critical thinking, science and logic as the best methods for evaluating all claims of fact, and we invite active participation by our members to create a skeptical community with a wide variety of viewpoints and expertise.


Home | Skeptic Forums | Skeptic Summary | The Kil Report | Creation/Evolution | Rationally Speaking | Skeptillaneous | About Skepticism | Fan Mail | Claims List | Calendar & Events | Skeptic Links | Book Reviews | Gift Shop | SFN on Facebook | Staff | Contact Us

Skeptic Friends Network
© 2008 Skeptic Friends Network Go To Top Of Page
This page was generated in 0.06 seconds.
Powered by @tomic Studio
Snitz Forums 2000