This comparison figure from the Minnesota workshop says it all. On one side, Prof. Park (upper image) continues to break ground with new data, and moves DIANA forward with full sincerity, humility, and transparency of a mature scientist; while on the other side, the petty old man (not shown here) is on a negative campaign to undermine the other's credibility with not only malice and selfish intent but also cowardice of hiding behind big names (such as shown here, Dr. Bandettini). I know full well the intention of both sides. Apart from science, this battle cannot be won by the petty old man on a purely moral ground.
This came across my computer this morning and I really appreciate it as I'm all about a healthy respectful dialogue where everyone has the best intentions. I certainly have respect for all involved and my co-authors on the paper I presented I have a respect grown from almost 30 years of appreciating their outstanding work. I would love DIANA to be more robust and reproducible - that is everyone's intent. So, my intention in my talk was to present what I knew and to clarify what was still a mystery. I should have made it clearer that, if the signal that one is looking for is consistent, then more averaging will lead to better detection. This I think everyone can agree on. As one of the questions rightly brought up, if the signal to be detected is not consistent and has an occurrence that doesn't follow a well characterized statistical distribution (i.e. more frequent early on and then dropping off in time), then finding a way to average only when it occurs is very very difficult, as it does risk starting a circular process where signal is artificially sculpted from noise. My goal was to start and continue a dialogue to address the central mystery that I attempted to present in my talk and which the Prof. Park alluded to but did not delve into detail about. How specifically were the "good" signals averaged when it was not known what their consistency was? This is a fundamental detail that I think at least one of the central points that needs to be better understood. It's an interesting problem and any information on it I think would truly help the international community better understand and replicate DIANA. I do have hope, and fundamental to advancing understanding is through sincere dialogue and mutual respect.
ReplyDeleteJust to reiterate. The key question that the entire brain mapping community has is: "How did you preselect your data?" In the spirit of transparency, it would be helpful to make your data available - both prior to pre-selection and after pre-selection. Your approach may have been fine, but it would be useful to share so that we can all learn something. So far, as I understand, these data have not been made available. Lastly, it should be noted, at the end of my talk, I welcomed discussion (as did the moderators), but Prof. Park did not chose to engage in any discussion at all.
ReplyDeleteThe "old petty man" published a paper titled "No Replication of DIANA" in Science Advances (https://www.science.org/doi/10.1126/sciadv.adl0999). I notified all co-authors, including yourself, about problems with data selection, but no one responded. The authors of the original DIANA paper ignored my requests to share their data and did not engage in any discussion (despite being in the same department). Let's keep our focus on the scientific aspects of this matter and refrain from resorting to personal attacks.
ReplyDelete