On 21 Aug 2003 20:09:39 -0700, [EMAIL PROTECTED] (Garbriel) wrote: > Thanks for your help, > You mean there is no approximation method, like adding 0.5 to each > cell, to do this case? > > If there is no varation in one varible. I want know that if there is > relation between these 2 variable, How can I do?
If you want to show that there is a relation between two variables, you have to KNOW that there is variation in both variables. A correlation is a statement about sample; you have to collect a sample that has variation. If you do have two ratings, say, that both doctors say that "Nobody has leprosy", you can say that they agree -- about this sample. And you hope that there is reason to say that "no leprosy" is reasonable, for this sample. If one doctor says, on the contrary, that "three patients do have leprosy", then you can comment on the general agreement between doctors; but now you have some disagreement evident: The dx made by one doctor lacks "sensitivity" or the dx made by the other lacks "specificity." You might test to say whether the excess count of differences is "statistically significant". You can't be sure about who is right unless you have some outside assessment that you want to call a "gold standard". You could try a google-search with some of these words. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html "Taxes are the price we pay for civilization." Justice Holmes. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
