What follows is a critique of the paper cited in another answer here: Setren, et. al., "Effects of the Flipped Classroom: Evidence from a Randomized Trial", Annenberg Institute at Brown University (2019). That paper doesn't, in fact, refute the value of the flipped classroom in any way and is an example of poor research design. In fact, it says nothing about a flipped classroom by any reasonable definition.
I'm adding this for the benefit of the O.P. who shouldn't be misled by "research" that purports to show that the flipped classroom is a poor educational practice, when it actually discusses something else entirely.
First, the design of the experiment in the paper by Setren et. al. (the authors) is fatally flawed. I would not have let any of my doctoral students continue with this research, though a number might have been interested in the topic. Fundamentally, if you can predict the outcome of a research project simply from the design, then it isn't really research. I don't doubt the sincerity of the authors but this seems like a situation of confirmation bias rather than a true search for truth regarding the flipped classroom.
The most serious issue with the paper, other than methodological failure, is an incorrect conception of the flipped classroom. The authors seem to have the mistaken impression that the "flipped classroom" is about where things take place. This is a serious misconception that prejudices any conclusion from the study.
In fact, the flipped classroom is about what takes place when the students and instructors are face to face. If they aren't different from the standard model, then nothing positive can be expected to happen. In a standard lecture the students are relatively passive, though they may take notes and, perhaps, ask questions. In a flipped classroom the students are active, typically in groups with collaborative educational activities intended to reinforce learning. Students come together in a flipped classroom to do things that can only be done face to face. Things that don't require interpersonal communication (not just broadcasting) are done otherwise.
In this experiment, the students were given videos to watch on their own, along with readings. No effective channel to ask questions is described, and neither are exercises based on that preparation. This alone is a flaw. It advantages people used to passively watching lectures, but it disadvantages those who need more help and have questions along the way. Separating the questions (next day in class) from the video is, itself a flaw. Students can, of course, make a written note of questions for answering later, and the best and most advantaged students are likely to do that, but only if they have enough experience with the format to recognize that it is essential. Students who try to rely on memory alone to bridge the gap between the video and the question period are disadvantaged. (See below for comments on lack of training.)
In most educational settings there is an expectation that students will spend twice as much time on a course outside the classroom as inside it. This seems to have been missing from the experimental setup as described. The videos were short, in particular (a good thing). But other preparation is also needed.
Also, active learning seems to have been actively discouraged. This again, advantages those students who don't need much instruction and are comfortable with relatively passive consumption of material. This is a poor way to generate insight, but most especially in those students who are otherwise disadvantaged by the educational system they have been faced with. The flipped classroom is actually an attempt to address this issue, hence group work. Group work is reported in the experiment, but mostly in the context of students solving problems, rather than more general learning and discussion. This again, seems to suggest a misconception. Students in a standard setting will, themselves, solve their exercises out of class partly in isolation and partly in a group context. But the flipped classroom, used properly, permits more opportunities since the instructor is present and can guide the activities.
The paper seems to suggest that video is a key to the flipped classroom. This is not actually true, though it is a common misconception. Students are given some way to prepare for class, but it needn't be videos. It could be text, such as PDFs prepared by the instructor. It could involve interactive simulations that let the students engage with the material. And it could (and should) involve some exercises based on that preparation.
The best flipped classrooms will also provide a way to get questions asked and answered before class, perhaps via a mailing list. Students might be required to write a one or two paragraph summary of the readings (or videos) as a sort of "ticket" for attendance in the subsequent face to face session. None of this was provided or expected in the experiment. With video alone, students are likely to be passive. But it need not be that way.
In a better course design, the flipped classroom will provide the instructor some feedback, prior to the class, that follows up on the out of class activities. If there is a mailing list, for example, the questions posed there are seen by other students and by the instructor. Even if they aren't answered on the list itself, the instructor can prepare the next class period in response to student concerns. The instructor, then, can spend less time and effort on things that the students already grasp well, and can focus on more essential concerns.
The paper seems to suggest that students work on exercises (alone or in groups) when in class with some chance for feedback, but I doubt that it is especially effective, just due to lack of time and lack of instructor resource. A sole instructor and a T.A. can only work with a few students at a time, leaving the other students bored and stuck. This is why pairing and group work is suggested, and normally required, for the flipped classroom. Students help one another with questions and explorations, not just exercises, freeing up instructors for the most difficult cases. Research has been done on this in the situation of paired programming in a communal space. The U.S.M.A. has strict ethical rules about joint work, though the paper doesn't say if these prohibitions against assistance were enforced here or not.
The faculty who actually carried out the experiment were all somewhat experienced in standard lecture formats and it was noted in the paper that they had little or no experience with other formats. Most of the instructors, however, were relatively inexperienced faculty members. As junior military officers, their own educational background was most likely on a standard model, with which they have excelled. Also, they had likely taught the target courses, or similar courses, in a standard format, similar to their own education. The experimental design doesn't describe that they were given any specific training in the flipped classroom. Nor were they permitted to deviate from the designed format. The paper does, however, admit that the students did better in those sections in which the instructor was more enthusiastic about the flipped classroom.
The paper also seems to admit that it may have been a flaw in the design that each instructor taught both a standard model section and a flipped section. Military officers follow orders, of course, but they don't have to accept the validity of what they are commanded to do. The whole process was, in fact, an imposition on them, as none were volunteers. This leads to uniformity of a certain kind, of course, but not to good educational outcomes for students.
Now, even if we somehow decide that all of this is to be forgiven, and that the flipped classroom model was well run (thus validating the data itself), there is still a fundamental flaw that would prevent us from drawing out generalized conclusions from the study: all of the subjects are from a highly advantaged subset of the US population. The U.S.M.A. is a highly selective undergraduate college in which students gain entry by proving that they excel in the standard educational process. They are used to lecture/exercise/exam based courses and have a history of success in such situations and most have little experience with alternate educational structures. Moreover, most of them come from dominant cultural and economic situations since admittance requires the recommendation of a member of Congress. They are not a cross section of students in any case, and especially, neither are their educational backgrounds.
Next, the experimental period was short, apparently between one and three weeks, with no one in the treatment group, student and instructor alike, given an opportunity to become comfortable or familiar with the process, while the control group had a long history of such experience with standard practice. Moreover, the authors report that students in the experimental group weren't very vigilant in doing the necessary preparation for class. This is likely due to their expectation from past experience that it was the classroom in which the material would be adequately "covered". But it was not. So, many seem to have entered the classroom unprepared.
It would be natural to expect a bit of chaos in the experimental group as the experiment was carried out in the middle of an otherwise standard course. All of a sudden, the rules of engagement are changed. This alone would lead one to suspect lower achievement over that lesson, as students have been dropped into a new world for a few weeks, not knowing how to act or what to expect, other than that they were supposed to "follow orders".
Given the experimental design, the outcome was foreordained. Put students into a new situation that might seem a bit chaotic, and they aren't likely to do very well.
Consider the following thought experiment in which I suspect that you can easily predict the outcome.
Suppose we had a large group of students who had spent their entire lives learning in a collaborative group environment in which they have access outside the classroom to a large variety of resources, but who spend their face to face time in collaborative learning. They can, and are encouraged to, prepare for class by reading, etc, and are also expected to bring questions with them. But there are also exercises they can do to test their own understanding prior to class. The face to face time is spent in reinforcement of the ideas and feedback on attempts. Students are encouraged to speak and offer ideas and other students are encouraged to comment.
Now, take the experiment as described in the paper, but flip the treatment and control groups. the course proceeds as normal for them (flipped) until we reach the start of the experiment. Otherwise run it as before, but in this case all of the students instructors are thoroughly familiar with the above described methods, but not so much with lecture to a relatively passive group, followed by exercises to be done at home.
What do you suppose the outcome of the new experiment would be. Isn't it pretty obvious?