 Hey everyone. Welcome back. It's Veronica. So we're going to continue talking about data analysis before we were talking about division. Now we're talking about stability. When we're talking about stability, what we're really looking for is how we reached what's called steady-state responding. And we want to make sure that for the most part that behavior is not improving on its own, that treatment effects are not temporary, and that we haven't intervened at the wrong time. We want to have a good picture of the phenomenon of interest of our client's behavior before we get started in treatment. Remember that we are talking about, just in a contextual analysis here, we're talking about the client behavior, we're talking about a child, 12 months, maybe they have some early signs of autism. We know that an early intervention is warranted, but who do we turn to for help? In our previous video, we looked at a couple of different interventionists and we were examining whether they were making a meaningful difference in the effect in the behavior. We were looking for division and neither of them did. Let's talk about what that would mean in terms of stability. So for Dr. C, this is the treatment effect that they've produced. We see a baseline condition. Remember on the y-axis, the left axis, that's our dependent measure, the number of smiles on the x-axis or the bottom horizontal axis. It tells us that we're looking at the number of smiles across time or sessions. And we have a baseline condition. We start with one smile, two smiles, three smiles, four smiles, and then five smiles. And in treatment, we jump up to seven smiles, eight smiles, six smiles, eight smiles, and seven smiles. Would you hire this doctor or would you look for someone else? I would not hire this doctor. And the reason that I wouldn't is because the data are trending towards treatment in the adjacent condition. So our baseline data are trending up. This is called an ascending baseline. It kind of looks like Annie's behavior, the number of smiles, is improving on its own. So if we have this ascending baseline, if the behavior is getting better on its own, why would we start treatment? How can we actually be sure if the therapist and not something else was what was making the difference in the behavior? These data are not stable because data in the baseline condition are trending toward or moving toward data in the treatment condition. So let's look at another doctor. Here's Dr. D. They come in for, again, same setup, the y-axis, number of smiles, x-axis, number of sessions. And we have in the baseline condition one smile, two smiles, two smiles, three smiles, and one smile. And then in the treatment condition, we pop all the way up to ten smiles, which is amazing, and then eight smiles, and seven, and six, and five. So would you hire this doctor? And again, I would not because the data here show me that that big improvement that we saw in treatment. It appears temporary. Those treatment data appear to be trending toward the lower level of data in the baseline condition. I would definitely not want to invest my time in a treatment that's only temporary. It means that I might be missing something really important. It means that I need to do better for my client. So just to quickly summarize, remember Miller says that stable data are those that are not moving closer to data in the adjacent condition. We're only ever going to compare conditions to one another one at a time. So baseline to reversal, or excuse me, baseline to treatment, treatment to reversal, or in the multiple baseline, we'll go baseline to treatment, baseline to treatment. We only ever compare one to another. And remember that data are plural. So data are data are in using Miller's approach to visual analysis, you only focus on the last three points of each condition. This is because treatment can sometimes take a moment to take a factor in the case like we saw with Dr. D. You can have a huge treatment effect, but then when you keep watching the data, they start to decrease because the treatment effect is temporary. So you want to make sure that you are waiting for the behavior to reach a kind of steady natural state before you make any condition changes. And then last, remember that when we're visually analyzing data, we do that by analyzing the last three of each condition. And we actually draw our trend line in Miller's approach to analysis by drawing from the first through the third of each condition the last three points. Check out the practice resources to see an example of what I'm talking about. Okay, so in the larger field, we would call this examining the trend in the data. But generally speaking in treatment and actual clinical practice, we're talking about the same thing is behavior getting better on its own. Our treatment affects only temporary. We want to make sure those behavior, we've got a steady state responding, they're not getting better on their own. There's not a ton of variability bouncing around and we want to make those decisions to change conditions to start treatment when we really understand the behavior. Check out the next series. We're going to be talking about determining whether your data are convincing or not. And then we're also going to talk about what role the actual design that you've selected demonstrates your experimental control and allows you to determine whether you have a causal relationship between your treatment and the behavior of interest that you're trying to change. I'll see you guys next time.