Comments - Difference Between Correlation and Regression in Statistics - Data Science Central2019-04-18T14:24:47Zhttps://www.datasciencecentral.com/profiles/comment/feed?attachedTo=6448529%3ABlogPost%3A702251&xn_auth=noActually Alok, Asim is correc…tag:www.datasciencecentral.com,2018-03-28:6448529:Comment:7077862018-03-28T20:18:42.572ZDan Butorovichhttps://www.datasciencecentral.com/profile/DanButorovich965
<p>Actually Alok, Asim is correct in his article. The correlation coefficient of your x and y is .975, I got the same result whether calculated by hand using the Pearson formula or calculated using R's cor(). Just because each y is a multiple or square of its corresponding x doesn't mean that it isn't estimable by a linear equation, or that they don't <strong>co-vary</strong>. In the case where you have truly nonlinear data, you can use other non-Pearson correlations such as Kendall's Tau, or…</p>
<p>Actually Alok, Asim is correct in his article. The correlation coefficient of your x and y is .975, I got the same result whether calculated by hand using the Pearson formula or calculated using R's cor(). Just because each y is a multiple or square of its corresponding x doesn't mean that it isn't estimable by a linear equation, or that they don't <strong>co-vary</strong>. In the case where you have truly nonlinear data, you can use other non-Pearson correlations such as Kendall's Tau, or Spearman's equations. Correlation is also about covariance, how much the two things vary together. As x changes, y changes and they do so together within the limits of the observation. Regression demands linearity, correlation less so as long as the two variables vary together to some measurable degree. </p> Hi Alok,
Very effective comme…tag:www.datasciencecentral.com,2018-03-28:6448529:Comment:7074982018-03-28T02:08:27.757ZAsim Janahttps://www.datasciencecentral.com/profile/AsimJana
<p>Hi Alok,</p>
<p>Very effective comment. See my below comments.</p>
<p>Two variables are said to be "correlated" or "associated" if knowing scores for one of them helps to predict scores for the other. Capacity to predict is measured by a <b>correlation coefficient</b> that can indicate some amount of relationship, no relationship, or some amount of inverse relationship between the variables.</p>
<p></p>
<p>From above comments, your point is correct. But your it can't be mixed up with…</p>
<p>Hi Alok,</p>
<p>Very effective comment. See my below comments.</p>
<p>Two variables are said to be "correlated" or "associated" if knowing scores for one of them helps to predict scores for the other. Capacity to predict is measured by a <b>correlation coefficient</b> that can indicate some amount of relationship, no relationship, or some amount of inverse relationship between the variables.</p>
<p></p>
<p>From above comments, your point is correct. But your it can't be mixed up with comparison.</p> "Correlation coefficient indi…tag:www.datasciencecentral.com,2018-03-22:6448529:Comment:7060952018-03-22T06:38:37.632ZAlok Kumarhttps://www.datasciencecentral.com/profile/AlokKumar252
<p>"Correlation coefficient indicates the extent to which two variables move together." - not really. <br></br><br></br><strong>Illustration</strong> - x (1,2,3,4,5,6,7,8, 9) and y (1,4,9,16,25,36,49,64,81) - x and y here move together. But what is the correlation coefficient? Even a Statistics Graduate passed out from the best of the colleges tend to say there is perfect correlation between the two. Actually not ! There is no correlation between the 2 variables. Don't you believe me? Calculate the…</p>
<p>"Correlation coefficient indicates the extent to which two variables move together." - not really. <br/><br/><strong>Illustration</strong> - x (1,2,3,4,5,6,7,8, 9) and y (1,4,9,16,25,36,49,64,81) - x and y here move together. But what is the correlation coefficient? Even a Statistics Graduate passed out from the best of the colleges tend to say there is perfect correlation between the two. Actually not ! There is no correlation between the 2 variables. Don't you believe me? Calculate the Corr Coeff and what you will get may surprise you.<br/><br/>Correlation Coefficient shows the extent to which they are "linearly" related ie the relationship between the two variables can be in expressed in the form of a straight line. Correlation is just a step on the way to regression.</p>
<p></p>