Co-eff of CORREL
Thanks Mike...I have a data with around 6 variables and I used one of them as
standard to find the relation with 5 others. now all the 5 variables had the coc in the range of 0.6 to 0.75. And there is high fluctuation in any one variable. Like, a single variable's relationship with the Standard Variable varies from 0.2 - 0.8 if relationship is calculated at different intervals of time (e.g. each day) what do you think should be the best way to reach at some pretty reasonable conclusion? "Mike Middleton" wrote: Malik - "It all depends." But here are some comments. First, Look At The Data. Create an XY (Scatter) chart. The correlation coefficient measures the extent of linear relationship. It's not appropriate for nonlinear relationships. Second, if you're a classical statistician, take the sample size into account. If you have a random sample and if you're really a frequentist, you can even perform a hypothesis test to investigate whether the observed correlation coefficient is significantly different from zero (or from some other value). Third, the value that is "taken seriously" varies from one discipline to another. In psychology, you might be satisfied with r = 0.3. In economics, you might want to see r = 0.9 or higher. - Mike Middleton http://www.DecisionToolworks.com Decision Analysis Add-ins for Excel "Malik" wrote in message ... Do u think co-eff of co-relation is reliable? I mean two variables would always show some relation between them no matter what. Also, at what value of COC should a variable be taken seriously? |
Co-eff of CORREL
Malik -
A "pretty reasonable conclusion" should depend on the context of your problem. You don't seem to have an Excel question. Instead, you may benefit from assistance by a statistician, data analyst, or decision analyst. - Mike "Malik" wrote in message ... Thanks Mike...I have a data with around 6 variables and I used one of them as standard to find the relation with 5 others. now all the 5 variables had the coc in the range of 0.6 to 0.75. And there is high fluctuation in any one variable. Like, a single variable's relationship with the Standard Variable varies from 0.2 - 0.8 if relationship is calculated at different intervals of time (e.g. each day) what do you think should be the best way to reach at some pretty reasonable conclusion? "Mike Middleton" wrote: Malik - "It all depends." But here are some comments. First, Look At The Data. Create an XY (Scatter) chart. The correlation coefficient measures the extent of linear relationship. It's not appropriate for nonlinear relationships. Second, if you're a classical statistician, take the sample size into account. If you have a random sample and if you're really a frequentist, you can even perform a hypothesis test to investigate whether the observed correlation coefficient is significantly different from zero (or from some other value). Third, the value that is "taken seriously" varies from one discipline to another. In psychology, you might be satisfied with r = 0.3. In economics, you might want to see r = 0.9 or higher. - Mike Middleton http://www.DecisionToolworks.com Decision Analysis Add-ins for Excel "Malik" wrote in message ... Do u think co-eff of co-relation is reliable? I mean two variables would always show some relation between them no matter what. Also, at what value of COC should a variable be taken seriously? |
All times are GMT +1. The time now is 02:07 PM. |
Powered by vBulletin® Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
ExcelBanter.com