Reading a book on historical methodologies, I noticed the inherent problem with making history scientific lies in the fact that history is not truly repeatable. Certain patterns repeat of course, but historical events are by definition a one-time affair.
Imagine a future historian with a time machine. Of course going back in a time machine to observe history has been the stuff of many stories, but imagine this future historian is more interested in the why behind past events. So he conducts changes in the timeline to see what happens as a result, making time in effect his experimental laboratory.
So, for example, say this historian takes on the question, "Was the transformation of the Roman Republic into the Roman Empire an inevitable consequence of a number events (such as increasing paralysis in the Roman senate)? Or was it because of the unique genius of Julius Caesar?" Historians of our day can read the evidence and make reasoned guesses, but this future historian can go back in time, assassinate Caesar while he is still in Gaul, and then move forward in this altered timeline to see if the Empire forms itself anyway. And if so, how.
Obviously a story featuring such historical methodology would presume that destiny is not fixed and the past could be changed, possibly through an application of quantum multiverse theory. Go back into your own past and change it, you have not eliminated your knowledge of what happened as would seem to be logically so--you have simply shifted out of your quantum multiverse into a different one, where a different set of quantum splits occurred, your own memories intact.
Probably a single historian would not be enough to carry out this task. More like an entire History Department, or an entire university would be engaged in repeatable and thus scientifically observable historiography. Once the "historical experiment" had been performed and the results noted, the above mentioned history professor could go back to before Caesar was assassinated for experimental purposes, stop the assassination, and return the timeline to normal.
As a story setting, this idea becomes much more interesting if something goes horribly wrong. Perhaps Napoleon or Ivan the Terrible capture the time machine and show an intuitive grasp of how it works and how to use it to dominate the world. Or perhaps, more subtly, the professor or team of historians the story focuses on gets lost in time somehow, having made changes they can no longer keep track of, now on a quest to set the timeline right--which of course has been done in science fiction many times before (setting the timeline right) most notably in the TV series Quantum Leap, though also Dr. Who at times and other books and shows have done this. Though I don't know of any who did so from the point of view of a history department trying to create "scientific history" through deliberate timeline changes.
What if a change in history was so subtle the historians could not trace how they had made it? Back to the example above, they kill Julius Caesar, the Roman Empire forms anyway, so they go back in time to restore Caesar's life, which they successfully do--but suddenly this restored man decides not to become emperor. And Rome remains a republic. And the historians, try as they might cannot figure out where they had gone wrong, not at all. Or else they discover they changed something they considered to be of no consequence that turned out to be huge, based on The Butterfly Effect (which is also the subject of a number of science fiction time travel stories)...