My Math Forum  

Go Back   My Math Forum > College Math Forum > Advanced Statistics

Advanced Statistics Advanced Probability and Statistics Math Forum


Thanks Tree1Thanks
  • 1 Post By v8archie
Reply
 
LinkBack Thread Tools Display Modes
November 3rd, 2014, 06:21 AM   #1
Senior Member
 
Joined: Oct 2013
From: New York, USA

Posts: 566
Thanks: 79

Is This Correlation Possible?

I made a regression equation with three independent variables and one dependent variable. Two of the independent variables had a positive coefficient (positive correlation) with the dependent variable and one independent variable had a negative coefficient. Then when I made a regression equation with that independent variable as the only independent variable, it had a positive coefficient with the same dependent variable. Is it possible for two variables to have a positive correlation on their own but appear to have a negative correlation when more independent variables are added in?
EvanJ is offline  
 
November 3rd, 2014, 07:53 AM   #2
Math Team
 
Joined: Dec 2013
From: Colombia

Posts: 7,042
Thanks: 2344

Math Focus: Mainly analysis and algebra
I'm reasonably sure that the answer is 'yes'. I think you might be talking about Simpson's paradox.

Last edited by v8archie; November 3rd, 2014 at 07:57 AM.
v8archie is offline  
November 3rd, 2014, 01:08 PM   #3
Senior Member
 
Joined: Oct 2013
From: New York, USA

Posts: 566
Thanks: 79

Quote:
Originally Posted by v8archie View Post
I'm reasonably sure that the answer is 'yes'. I think you might be talking about Simpson's paradox.
I don't think my question is in that category. Simpson's Paradox involves two variable individual vs. combined, with the same amount of variables examined both times. I have one equation with four variables and one equation with two variables. Call them a, b, c, and x, with three independent variables and x being the dependent variable.

Equation 1 used every variable and:
a and x had a positive correlation
b and x had a negative correlation
c and x had a positive correlation

Equation 2 had only b and x and in that case b and x had a positive correlation. Is that possible?
EvanJ is offline  
November 3rd, 2014, 02:15 PM   #4
Math Team
 
Joined: Dec 2013
From: Colombia

Posts: 7,042
Thanks: 2344

Math Focus: Mainly analysis and algebra
Maybe the same amount of variables each time, butthe different environments equate to other variables having differing values, perhaps uncontrolled, or perhaps controlled unintentionally.
Thanks from EvanJ
v8archie is offline  
June 9th, 2017, 06:20 PM   #5
Senior Member
 
Joined: Oct 2013
From: New York, USA

Posts: 566
Thanks: 79

Quote:
Originally Posted by v8archie View Post
I'm reasonably sure that the answer is 'yes'. I think you might be talking about Simpson's paradox.
Over 2.5 years later, I looked farther down that page after skipjack linked to it another topic. What I said is a type of Simpson's paradox.
EvanJ is offline  
Reply

  My Math Forum > College Math Forum > Advanced Statistics

Tags
correlation



Thread Tools
Display Modes


Similar Threads
Thread Thread Starter Forum Replies Last Post
Correlation A and B Necessity Advanced Statistics 2 August 21st, 2014 11:46 AM
correlation najaa Algebra 2 April 27th, 2012 02:18 AM
correlation coefficient matmul Advanced Statistics 1 October 24th, 2011 09:11 AM
JPD's & Correlation fin0c Advanced Statistics 4 January 24th, 2011 08:25 PM
MIN, MAX from correlation daniel2 Algebra 3 October 5th, 2009 12:55 AM





Copyright © 2017 My Math Forum. All rights reserved.