deletions | additions       diff --git a/In_cite_streitburger2014impact_the_researchers__.tex b/In_cite_streitburger2014impact_the_researchers__.tex  index e113aca..e67732f 100644  --- a/In_cite_streitburger2014impact_the_researchers__.tex  +++ b/In_cite_streitburger2014impact_the_researchers__.tex 
...
In \cite{streitburger2014impact}, the researchers found that different pulse sequences and hardware (12 ch vs 32ch RF coil) altered the estimate of the coefficient on gray matter density in an age regression. This is what we are hypothesizing for regional volumes as well. $D_{u,j}$ is the \textit{unobserved} effect - in the case of \cite{streitburger2014impact}, the true GM density dependence on age that cannot be measured directly. We propose that the reason for the differences seen in regional volume analyses (and possibly VBM) is due to different scaling biases, $a_j$, for different pulse sequences. So it is definitely true that $a_j$ and $D_{Y,j}$(the \textit{observed} effect) are correlated, but we do not think there is a correlation between the \textit{unobserved} effect ($D_{U,j}$) with $a_j$, which is only dependent on scanner hardware/acquisition. To verify this assumption, we calculated scaling factors on MS patients and healthy controls and could not detect differences between them.