Comments on: The Latest Chapter in the Great AVE Debate http://bynd.com/2010/01/21/the-latest-chapter-in-the-great-advertisting-value-equivalency-debate/ rethinking digital Mon, 02 Jan 2012 15:56:02 +0000 hourly 1 http://wordpress.org/?v=3.2.1 By: Seth Duncan http://bynd.com/2010/01/21/the-latest-chapter-in-the-great-advertisting-value-equivalency-debate/#comment-87 Seth Duncan Sat, 30 Jan 2010 00:13:48 +0000 http://context-analytics.com/?p=361#comment-87 Thanks for the thoughtful comment, Brad! Definitely keep me posted if you run a multiple regression analysis on all metrics included in the weighted media cost composite. It would be interesting to see the results. Thanks for the thoughtful comment, Brad! Definitely keep me posted if you run a multiple regression analysis on all metrics included in the weighted media cost composite. It would be interesting to see the results.

]]> By: Anonymous http://bynd.com/2010/01/21/the-latest-chapter-in-the-great-advertisting-value-equivalency-debate/#comment-86 Anonymous Thu, 28 Jan 2010 15:20:32 +0000 http://context-analytics.com/?p=361#comment-86 Seth, You are right that a multiple regression would in fact be a better test to determine which of the factors best predicts the business outcomes, but that wasn't the claim of the paper (as you correctly noted in your blog). The paper isn't arguing that Weighted Media Cost is the best predictor of the business outcome, but that including it in a mixed model increased the correlations and coefficients explaining the increase in the outcome. In each case where we included WMC it improved the correlation and coefficient. As you said, it is somewhat of a no-brainer that if you include more variables in your model you are likely to improve the model. However, this is only true if you are adding valid variables. Despite all of the measurement jargon in the paper, it is really proposing a simple idea: If you include data that evaluates the quality of the medium where the message appears (which, arguably, advertising costs does) then it provides a better measure of the effectiveness of the message. I think it would be interesting to run a multiple regression with the model to determine which factor best predicts the outcome, and we should look at this as the next step in this line of inquiry. Perhaps we can publish another paper on this subject and you'll blog about it again. Brad Seth,

You are right that a multiple regression would in fact be a better test to determine which of the factors best predicts the business outcomes, but that wasn’t the claim of the paper (as you correctly noted in your blog). The paper isn’t arguing that Weighted Media Cost is the best predictor of the business outcome, but that including it in a mixed model increased the correlations and coefficients explaining the increase in the outcome. In each case where we included WMC it improved the correlation and coefficient.

As you said, it is somewhat of a no-brainer that if you include more variables in your model you are likely to improve the model. However, this is only true if you are adding valid variables. Despite all of the measurement jargon in the paper, it is really proposing a simple idea: If you include data that evaluates the quality of the medium where the message appears (which, arguably, advertising costs does) then it provides a better measure of the effectiveness of the message.

I think it would be interesting to run a multiple regression with the model to determine which factor best predicts the outcome, and we should look at this as the next step in this line of inquiry. Perhaps we can publish another paper on this subject and you’ll blog about it again.

Brad

]]>