<p>hawkette-
I read the article linked in your post #61. The author is trying to make a case for enhancing graduate programs and faculty research at Emory. The article was published on the Academic Exchange, which, I believe, is not a refereed journal reviewed by professionals.</p>
<p>The author did basically the same thing I did. He correlated peer assessment with measures of graduate program strength and compared the correlations with measures of undergraduate program strength. He concluded that the correlations between peer assessment and graduate/research strength were higher than correlations between peer assessment and measures of undergraduate strength.</p>
<p>However, he is wrong about this. He is incorrect. My analysis showed that peer assessment was very highly correlated with measures of undergraduate strength. The author said he didn’t find any correlations between PA and undergraduate strength higher than .67 but I found a correlation between PA and SAT of .79.</p>
<p>The author also commits a fundamental error in the conclusions he drew. He suggested that his correlations implied a causal relationship between PA and graduate ranking but he can’t draw that conclusion. It could simply be that PA is an index of undergraduate strength but graduate/research strength and undergraduate strength tend to occur together. So, PA could reflect undergrad strength but grad/research strength is tied to undergrad strength. Grad and undergrad strength are inseparable.</p>
<p>His conclusions clearly show a lack of objectivity, lack of research skill, lack of understanding of the scientific method. He was biased.</p>
<p>Why did he fail to find a higher correlation between PA and measures of undergrad strength? It could be due to an error in statistical methodology known as “truncated range”. He only chose the top 30 schools whereas I used the top 100 or so. Truncated range can result in underestimated correlations.</p>
<p>Why did he find such a high correlation between PA and graduate program rank? As I recall, the author used a sort of composite measure of graduate program rank and obtained a correlation of about .83. If I had used a composite measure of undergraduate program strength, my correlation between PA and a composite measure of undergrad program strength would be about .9.</p>