Notices
Results 1 to 3 of 3

Thread: Chi-squared question

  1. #1 Chi-squared question 
    Forum Radioactive Isotope MagiMaster's Avatar
    Join Date
    Jul 2006
    Posts
    3,440
    I was wondering about the chi-squared test (specifically, Pearson's chi-squared test). Would it be more informative to perform the test on all n datapoints at once, or perform k seperate tests on each set of n/k datapoints.

    An example would be testing a (standard 6 sided) die to see if it's fair. You roll it 600 times, storing each roll. Would you get more information about the fairness of the die from looking at 10 sets of 60 rolls, or 1 set of 600 rolls?

    What if, while looking at the results from 10 sets, 4 of those gave a result between .05 and .1 (not low enough to generally be considered a failure on their own, but close)? Would this necessarily imply anything (either about the die or about the results of looking at all rolls)?


    Reply With Quote  
     

  2.  
     

  3. #2  
    Forum Radioactive Isotope MagiMaster's Avatar
    Join Date
    Jul 2006
    Posts
    3,440
    Is it a good sign when my questions don't have an easy answer? :wink:


    Reply With Quote  
     

  4. #3 Re: Chi-squared question 
    Suspended
    Join Date
    Apr 2008
    Posts
    2,176
    Quote Originally Posted by MagiMaster
    I was wondering about the chi-squared test (specifically, Pearson's chi-squared test). Would it be more informative to perform the test on all n datapoints at once, or perform k seperate tests on each set of n/k datapoints.

    An example would be testing a (standard 6 sided) die to see if it's fair. You roll it 600 times, storing each roll. Would you get more information about the fairness of the die from looking at 10 sets of 60 rolls, or 1 set of 600 rolls?

    What if, while looking at the results from 10 sets, 4 of those gave a result between .05 and .1 (not low enough to generally be considered a failure on their own, but close)? Would this necessarily imply anything (either about the die or about the results of looking at all rolls)?
    To be honest when you start using statistics you or someone else already knows the answer. But rather then do something about it, you try to persuade the populous with your statistics.

    I mean look, statistics are to show something. And if you already have a clue, you are creating the statistics to show something, that you already know about.

    So your question is a bit out of order. If you are trying to show that errors occur in groups. Someone must want to overlook this or else they would be right on the problem already.

    I have worked with people and systems like this. Statistics tend to make phony jobs, but usually show or do little.

    Besides real statistics are thrown away and phony statistics are held high.


    Sincerely,


    William McCormick
    Reply With Quote  
     

Bookmarks
Bookmarks
Posting Permissions
  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •