Okay, In this video, we're going to address this question. There's a question that I get asked a lot when I'm teaching data transformation and students will ask me, is transforming data? Cheating? Answer is no. And the most fundamental reason for why transforming data is not cheating is that we treat all of our data points equally. And when we treat all the data points equally, we avoid adding bias to our analysis. Having said that, we do need to be really careful when we're transforming our data. Which is something that I highlighted at the end of the previous video. When we're choosing a transformation, we want to choose a transformation that can effectively allow our data to meet the assumptions of the test that we're doing. When we do that than transforming our data allows us to produce results that will be more reliable than if we had not transform the data. If, however, we were to choose a transformation not based on how well our data meet the assumptions, but instead on some aspect of conclusions. Like if we chose a data transformation based on the p-values that we got, that would be cheating and that is something we definitely do not want to do. So just to reiterate, the point I just made, the whole motivation for transforming data is to try to make sure that we analyze our data in a way that gives us the most reliable results possible. That's our goal. And so our goal of transforming data is the opposite of cheating. We're trying to do something to allow us to report reliable results. The last thing that I want to mention is that often the scale in which you measure things is fairly arbitrary anyways. And I want you to consider just a simple example of like enzyme abundance. We might be interested in how enzyme abundance. We might be interested in measuring the enzyme abundance in cells in a number of different contexts. So for example, we might be interested in the abundance of a particular enzyme for one genotype versus another. And we might be doing that. We might be interested in enzyme abundance because we might think that the abundance of the enzyme will tell us something about how effectively a particular pathway might function. But we know from studies of biochemical pathways that the abundance of an enzyme does not translate directly into the rate at which a particular pathway will function. If you decrease the abundance of an enzyme in half. And that's unlikely to decrease the overall flux of a pathway by half as well. The point in trying to make here is that we often measure variables in biology in a way that we hope will be meaningful to understanding the underlying biology. But the scale that we've used to measure things may not always be entirely clear. And often the scale that we choose to measure things can be arbitrary. And so if the scale that we're using is arbitrary to begin with, and I'm not seeing it always is. But often the scale at which we use to measure something is arbitrary. And those cases y naught, transform the data. Why not change one arbitrary scale to another arbitrary scale? The last thing I want to end on is this question of whether or not data transformation changes our interpretation of the results. There's a lot on this slide. I'm going to go through it quickly. The short answer is, well, that depends, which is what I have listed here first. So first of all, the interpretation of p-values does not change. But our interpretation of effect sizes can change. And we might want to do something called back transformation. If our model includes interactions, which is something that we'll discuss when we talk about two-factor general linear models or analysis of covariance, etcetera. And those circumstances, interpretations of interactions in those and those analyses can be affected by data transformation. Also, if we have continuous covariates. In other words, if we have an independent variable in our analysis that is not a factor or doesn't have just a series of categories, but it's continually distributed. Than transforming the data could change your interpretation of the effect of that continuous variable. The point here is that if we transform our data, then we sometimes are going to have to accommodate how we interpret the results that we report. And we want to make sure that we interpret them in a way that's accurate given the transformation that we may have performed. But I'd argue that it's better to do that. And to do that for analysis that is based upon data that do meet the assumptions the test, then to keep our interpretation simpler and report results that are not necessarily reliable. And we'll end there and say, I hope that this brief discussion of cheating, it's been helpful. Thank you.