How to choose the scrubbing method in DPARSFA?

Submitted by Yangmei Luo on Mon, 08/03/2015 - 09:50

Dear Dr. Yan,

How to choose the way to scrub method in DPARSFA? Could you give me some suggestions on this issue?

My reviewer asked me to scrub the bad data on a RS-fMRI paper. I used DPARSFA to do this following the reviewer. However, he said “the scrubbing method implemented in DPARSF only adds a regressor for the time point in question and is less rigorous than that described in Power et al 2012 or 2014, which removes the time points and adds synthetic data”.

I think the reviewer may think I used the top “scrubbing” button in DPARSFA. In fact, I used the bottom “scrubbing”, which have multiple ways of interpolation, such as “cut”, “Nearest Neighbour”, “Linear”, “Cubic Spline”, “Piecewise Cubic Hermite”. However, which methods is mostly closed to Power et al 2012 or 2014? I know you are one of the expert of head motion, could you give me any insight?

Thanks a lot in advance!

Best!

Yangmei

Forums:

YAN Chao-Gan

Tue, 08/04/2015 - 02:45

Hi Yangmei,

If you used the bottom "scubbing" with "cut", it's Power et al., 2012 method.

Best,

Chao-Gan

Yangmei Luo

Tue, 08/04/2015 - 15:14

Hi Chaogan,

Thank you for your quick response! I appreciate your kind help!

Does the "cut" method mean to remove the bad time point totally? But Power (2014, Neuroimage)  seem to use the good data replace the bad data. I am very worried that the "cut" method may change my main previous findings, since it's a group ICA study. In previous scubbing, I used the “Nearest Neighbour” method, which seems to do not change much on the results. Could you give me a suggetion on this question?

In addition, could you explain the methods to scubbing in more detail? They include “cut”, “Nearest Neighbour”, “Linear”, “Cubic Spline”, “Piecewise Cubic Hermite”? 

Thanks  a lot in advance!

Best,

Yangmei 

YAN Chao-Gan

Thu, 08/06/2015 - 03:48

Hi Yangmei,

"cut" means to remove the bad time point totally, that's Power et al., 2012.

For replacing, there are several interpolation methods: “Nearest Neighbour”, “Linear”, “Cubic Spline”, “Piecewise Cubic Hermite”, from simple to complicated. Personally I will use Linear.

Best,

Chao-Gan