Okay, we've now had a series of videos where we've done some pretty heavy stuff. We've talked about what non-independence is or what independence is with respect to data. We've talked about how non-independence can leads to pseudo replication and thereby really interfere with our ability to make proper conclusions from experiments. And we talked about how non-independence can creep into our experiments in ways that we might, might not normally think about. In this video, we're gonna do something very different. This video is all going to be a series of challenges for you. Where what I'm gonna do is I'm going to show you some results published in a series of papers. And I'm going to challenge you to decide whether or not there's potential or evidence for pseudo replication in that study. Okay? And we're going to start with this experiment, where this is a cartoon sketch of an experiment that aims to answer this question. What is the effect of autism associated and L3 mutations on synapse function at the Catholics of Held. I don't know what that means. I'm not that kind of biologist. I'm a biologist, but this is outside my area of expertise exactly understanding what those words mean. But that doesn't matter. It doesn't matter. What does matter is our ability to look at how an experiment is designed. And to look at how the, the, the experiment was analyzed to determine whether or not the experiment and its analysis were appropriate for a general purpose. Okay, so what I've done here is I've created a cartoon sketch of this experiments that address this question. And so what this experiment uses a series of mice where within, for each mouse, they took a series of cells and then they measure the qualities of each of those cells. So I have The way I picture this is I'm saying, let's imagine we took three cells from each individual mouse. I don't know whether or not it was always exactly three cells. The point is here, we've got multiple cells from each mouse, OK. And we have some mice that were allocated to a control treatment and some mice that were allocated to a mutant treatment. And the study analyzed the responses of the individual cells. So I want to ask you stop and think about this for a minute. Could this study be pseudo replicated? Okay, now that you've had a think, let's move on. Okay. First of all, let's ask how could we be certain to avoid pseudo replication? Well, one way would be to not analyze the data coming from each individual cell. Because these will not be independent, or at least will have good reason to believe that they are not independent. And that's because they all come from the same individual. This is essentially measuring the same individual more than once. And so what we could do to avoid this, what we could do to make sure that any analysis of our data did not involve SR replication. Excuse me. Assuming that all other aspects of the experiment did not introduce sources of non-independence. What we could do is we could just calculate the average value or the average quality of the cells from a mouse in order to have one observation per mouse. So what that would mean from how I've drawn this here is that we would have two observations for the control group and two observations for the mutant group. And then we can analyze and mean response per mouse. There are fancier methods as well that can account for potential non-independence due to things like this. In later videos, we'll talk about how we can use mixed effects models in order to account for these kinds of scenarios. Okay? So I asked you though whether or not these data you thought could be pseudo replicated. Let's see what the authors did. Well, the way the authors analyzed the data did introduce pseudo replication. We can see this from looking at this figure. Where in the figure they note the number of cells that were used for a particularly treatment and the number of mice that were used. And there's no other indication in this paper that they did anything to account for the fact that they had multiple measurements per mouse. So these results will come from analysis that allowed pseudo replication to occur. Let's know some more details about this study. Not, not to pick on anyone, but I think I will just to make a larger point. So this, here's, here's a study that these data came from. And one of the things that I want to point out is that the lead author or the P, the PI or principle investigator, is actually Nobel Prize winner. And I'm not. If, if if this person ever hears about this video in any way, please rest assured. I I'm not trying to pick on any particular individual. I'm trying to make a larger point. And the larger point is, no matter how accomplished someone is, no matter how intelligent people are, mistakes can happen. And so whenever you're looking at a study, you always have to look at it with fresh eyes. Don't worry about who did the actual experiment, who was involved. Don't worry about where it was published. Look at it with your own eyes. Because it's possible that mistakes have creeped in as I have in this study. Okay. So that's one example. I want you to see if you can spot pseudo replication or evidence for studio replication. In this output, are this, this snippet that I've taken from another paper? Just pause the video for a moment, read what you've got here and see if you can identify it. Okay? Now that you've had a chance to think about that, what you can see here is we can see that the control has an N of 50. So Anna, 50 neurons. But those 50 neurons only came from three brains. And again, for the mutant, there were 71 neurons, but they only came from for brains. And again, there's no indication in this study that further measures were taken to account for the non-independence. It's very likely to be occurring here. This paper was published. Here. It was published Nature. Which is a very big journal, journal that attracts lots of attention. Again, one of the, one of the things that I really want you to appreciate is that just because something is published in a famous journal like this does not mean that you can trust the results. You always have to look at things with fresh eyes based on your own experience and your own training. What about this example? You get the idea now, I want you to pause the video and look what we have here, okay? When you're ready, just start the video up again. Okay, now that you've had a chance to look, okay, which you can see here is says number in parentheses represents a number of slices. So we can see here we have 11 slices for wild-type, 12 for the other group, 16910, 12s, we have lots of slices and each of these, and then it says at least three animals per genotype, per age were used. So with out further information, it certainly looks like they're strong potential for pseudo replication associated with these results. Again, just because if we're taking, if we have many more slices, then we have animals than that suggests that we're going to have non-independent data that came from our particular animals. And once again, this paper came from, or these data came from this paper which is published in cells, which is one of the huge journals in the field of biology. So once again, our lesson is, or the less that I'm really trying to emphasize. Don't worry about where something's published when you're reading a paper and deciding for yourself. Whether or not you believe what's there. Don't pay attention to the fact that it's in one journal or another or who the authors are. The best thing that you can do is to look at the results with your own fresh eyes and using your own experience and training. Here's just a last example. It's not drawn for any particular paper, but it's just a common example that I just wants to really draw to your attention. Let's imagine that we had an experiment. It was set up like this, where we had a series of families. So here's a series of mothers. And then from each mother We had a series of offspring. And then we analyzed these offspring. And let's say that we had multiple offspring from a particular mother allocated to a particular treatment group. Ok, so that follows from the ideas that we discussed in some previous videos about how non-independence is especially worrying if we have non-independence within treatment groups. So in this case, we would really want to be wary of how we would analyze our data to make sure that we can account for the potential of four potential influences from non-independence. And we could ask how many independent mice do we have among the offspring? Well, we have eight offspring, but it's really debatable exactly how independent they are. At the least, we could say that we have for independent units, okay, well, we could argue that each family is likely independent, ok. But the data within the families are, we have good reason to think that they would not be independent. So how would we analyze data from an experiment like this? Where if, if we had all these data within one particular treatment, well, there's at least two ways we could approach this. One would be to, again, take the mean of the mean value among our mice from a particular mother. And so, and that way we collapse a larger number of data points into a single independent data point. So that would be one thing we can analyze family means. Or alternatively, we could use a particular statistical tools that allow us to account for possible sources of non-independence. And again, we're going to talk about mixed effects models as one method to do that. And we're going to discuss that in later videos. And I'll stop there. I hope that this video has been a little bit entertaining. What really, what I'm trying to do is to give you some practice to train your mind, to train your eyes, to spot and also to avoid pseudo replication in experiments that you read about. And when you're designing and conducting your own experiments. And with that, I will say, thank you very much, and I hope this video has been helpful.