Latest / This Is Why with Dr. Busti / Intro to Biostatistics: How to Pick the Right Statistical Test

Transcript

Intro to Biostatistics: How to Pick the Right Statistical Test · This Is Why with Dr. Busti

• 0:00
  SPEAKER_00: When you hear the word biostatistics, what goes through your mind?
• 0:07
  SPEAKER_00: Sometimes people are very intimidated by this subject and that term.
• 0:12
  SPEAKER_00: Some people get PTSD from it, some people kind of get excited about it, right?
• 0:18
  SPEAKER_00: That's okay.
• 0:21
  SPEAKER_00: We all are in the same boat, and we all are expected to be able to understand how to conduct clinical research at some basic level in its framework.
• 0:32
  SPEAKER_00: But as clinicians, also we need to understand when we're reading a paper, what was the researcher supposed to do, or what would be the best way to organize a study and develop a protocol, carry out the study so that it can be analyzed appropriately, because at the end of the day, we need to be able to interpret that literature and translate it to the bedside, the patient sitting in front of us.
• 1:00
  SPEAKER_00: And so biostatistics becomes very important in understanding what the results actually mean for the sample of the population that was studied, so that we can then determine if it's internally valid and extrapolate it back to the population.
• 1:16
  SPEAKER_00: Now, guys, listen, I do not pretend and I will never claim to be a biostatistician.
• 1:22
  SPEAKER_00: I don't think you need to be one to do clinical search.
• 1:26
  SPEAKER_00: I've published many papers and have done conducted clinical research, but I've also been that active clinician still to this day who has to read the papers, interpret the literature, and apply it to my patients that I am also serving.
• 1:41
  SPEAKER_00: So that is the goal here in this session, which is an introduction to biostatistics.
• 1:49
  SPEAKER_00: Getting us oriented is critical to whether you're developing a research protocol and conducting research or you're that clinician reading the paper.
• 1:58
  SPEAKER_00: All right.
• 1:59
  SPEAKER_00: So with that, I'm well I'm Dr.
• 2:00
  SPEAKER_00: Bustaye and welcome to This Is Why, a show that is dedicated to helping all of us understand why we do what we do because it makes the how much easier.
• 2:10
  SPEAKER_00: And the how here is how to develop a research protocol appropriately, how to conduct the study, how to analyze it and publish it, but also if you're the clinician, how to read and properly interpret the paper so you can apply it to the patient in front of you.
• 2:26
  SPEAKER_00: Okay.
• 2:27
  SPEAKER_00: And on that note, we're gonna go ahead and start it.
• 2:29
  SPEAKER_00: And I want to get you oriented to the series.
• 2:32
  SPEAKER_00: Okay, there's three different parts to this series, and and while the individual topics can stand alone, there is at some level a little bit of an order.
• 2:45
  SPEAKER_00: Okay, so if you need that, we put those together for you in our in a collections or playlist that are available not only on the YouTube channel, also on our podcast channels, but also at thisiswhy.health to make it easy for you to navigate through the steps of doing a clinical search and the protocol development or how to properly interpret the paper.
• 3:11
  SPEAKER_00: Okay, so part one is you actually have to consider statistical analysis before the study starts, and that's part of the statistical analysis plan or the SAP.
• 3:23
  SPEAKER_00: Okay, part two of this sort of module series is the attempt at the completion of the study, you now need to analyze it and produce some results so that you can interpret, put some meaning on that.
• 3:38
  SPEAKER_00: And that is considered summary statistics, and it's broken down into descriptive versus inferential statistics.
• 3:46
  SPEAKER_00: And then lastly is diagnostic tests, which is kind of a you know standalone by itself where we do measures of accuracy and performance if the study has to do with that.
• 3:57
  SPEAKER_00: Okay, now what I put together for you is a more detailed outline so that as you engage with this, if you want that orientation and framework, this is where we're going.
• 4:08
  SPEAKER_00: So here we are on the statistical analysis plan, and we're at the introduction of biostatistics, getting oriented.
• 4:15
  SPEAKER_00: That's this lecture right now.
• 4:17
  SPEAKER_00: Following it will be what type of data variables are you dealing with?
• 4:23
  SPEAKER_00: Is that nominal data, ordinal data, continuous, and how do you determine?
• 4:27
  SPEAKER_00: Okay, and then the type of study groups.
• 4:30
  SPEAKER_00: Are you dealing with independent, related?
• 4:33
  SPEAKER_00: How many of them?
• 4:35
  SPEAKER_00: One, two, three or more, okay, or two or more, right?
• 4:39
  SPEAKER_00: All of those things become relevant when you're considering the type of statistical analysis that needs to be done.
• 4:45
  SPEAKER_00: And that is the standardized traditional parametric versus non-parametric statistical analysis.
• 4:51
  SPEAKER_00: And again, as some of you may be already starting, they're like, oh my God, right?
• 4:55
  SPEAKER_00: I'm gonna make this easy for you at some level, okay?
• 4:59
  SPEAKER_00: I mean, I've been studying this, I've been teaching this for over two decades.
• 5:04
  SPEAKER_00: And over time, I felt like it was difficult early on, but as I began to work through it, I began to find ways to make it a little bit easier to understand.
• 5:15
  SPEAKER_00: And that's my goal is to translate that back to you.
• 5:18
  SPEAKER_00: We'll also dive into some other additional things of study design like alpha and power analysis.
• 5:23
  SPEAKER_00: All of these things must be considered before a study is started.
• 5:28
  SPEAKER_00: That's the appropriate way.
• 5:30
  SPEAKER_00: And many times when you're going through the IRB, if you're on the side of developing clinical research or you're learning how to do that and developing a protocol, the IRB is going to expect that.
• 5:40
  SPEAKER_00: They want to make sure you know what you're doing and what you're going to analyze and how you're going to analyze it so that when you produce the results, they're meaningful for the reader.
• 5:49
  SPEAKER_00: Okay.
• 5:49
  SPEAKER_00: Then you complete the study and we go into part two.
• 5:52
  SPEAKER_00: And again, that summary statistics, which is that descriptive versus inferential statistical analysis, where we learn how to describe the data.
• 6:02
  SPEAKER_00: Okay, and what does that mean and what does that look like?
• 6:05
  SPEAKER_00: The central tendency, the variability of the data variables that we're getting, how dispersed are they in the sample?
• 6:12
  SPEAKER_00: Measures of association or correlation, okay?
• 6:16
  SPEAKER_00: Doing measures of relative dispersion and precision, okay?
• 6:20
  SPEAKER_00: Those are like coefficients of variations when it's applicable.
• 6:23
  SPEAKER_00: Inferential statistics are those common metrics or variables that we think about p-values, 95% confidence intervals, you know, the effect size, things like odds ratios, relative risk, hazard ratios, number to retreat, number needed to harm, when do I do a regression analysis?
• 6:40
  SPEAKER_00: Okay, all of those things become very relevant in our ability to take the data from that kind of study and that study design, okay, which we've covered in a previous lecture, if you need to review that, helps to drive some of these decision points.
• 6:58
  SPEAKER_00: And then in the diagnostic test, if it were a diagnostic related study, again, that accuracy precision, we're thinking about things like precision, accuracy, sensitivity specificity, positive and negative predictive values.
• 7:10
  SPEAKER_00: How do you apply that?
• 7:11
  SPEAKER_00: What does it mean?
• 7:12
  SPEAKER_00: What does it not mean?
• 7:14
  SPEAKER_00: What is an ROC curve, right?
• 7:17
  SPEAKER_00: Likelihood ratio so that we have a pretest probability.
• 7:20
  SPEAKER_00: What is the likelihood that our post-test probability is changed from doing something, some diagnostic test?
• 7:29
  SPEAKER_00: Okay, so that is the overview of the series at a high level broken down into subcategories as sections, okay?
• 7:41
  SPEAKER_00: Now, when we think about the steps of clinical research, you guys have seen this when in this particular collection on clinical research, evidence-based medicine, biostatistical interpretation, and clinical and protocol development, all that stuff that we're talking about.
• 7:56
  SPEAKER_00: I keep coming back to this because it is the place of orientation.
• 8:01
  SPEAKER_00: It's not meant to insult you or waste time or drag this out.
• 8:05
  SPEAKER_00: But we have to stay oriented at where are we at in the process of clinical research.
• 8:09
  SPEAKER_00: And again, if you're not the researcher, it's okay.
• 8:12
  SPEAKER_00: As a reader, you still need to envision as you read the protocol what should have been going on at this step as you read through the methodologies that were given to you.
• 8:24
  SPEAKER_00: And you should be able to replicate someone's study.
• 8:27
  SPEAKER_00: If they give you a proper publication, you should be able to know exactly what to do.
• 8:32
  SPEAKER_00: So we have that developing the study concept, you secure funding, you prepare the study protocol.
• 8:39
  SPEAKER_00: Okay.
• 8:40
  SPEAKER_00: You obviously then enroll once you're approved, you implement, you collect the data, you finish it, you close the study, you then analyze it.
• 8:48
  SPEAKER_00: So here are the two areas that we are going to be anchoring in on this particular module or series.
• 8:56
  SPEAKER_00: Again, the development of the study concept up here at the very top.
• 9:01
  SPEAKER_00: And when you're developing the study protocol, you can see one of the core content areas is the consideration of your basic statistical decision tree.
• 9:13
  SPEAKER_00: And that's what that SAP is really trying to guide you through.
• 9:17
  SPEAKER_00: How are you, based on the study design, the hypothesis?
• 9:22
  SPEAKER_00: What is the hypothesis's endpoint?
• 9:25
  SPEAKER_00: What kind of data is that?
• 9:27
  SPEAKER_00: What kind of variable is it?
• 9:29
  SPEAKER_00: Because that drives everything that you're going to do in picking what statistical test needs to be done to analyze that result.
• 9:39
  SPEAKER_00: And then also you need to consider ways to improve the internal validity, right?
• 9:45
  SPEAKER_00: You don't want to make a type 1 or type 2 error in the results.
• 9:50
  SPEAKER_00: And so the researcher's job is to appropriately do this before, and that's why the IRB is in place in part, is not only to protect the participants from the researcher doing something that's not right or unethical, but it's also making sure that researchers equipped with the team members.
• 10:09
  SPEAKER_00: Maybe he or she doesn't necessarily know how to do the statistical analysis.
• 10:13
  SPEAKER_00: Again, I'm not a biostatistician, don't pretend to be one, don't want to be one, right?
• 10:18
  SPEAKER_00: But I need to understand that role even as a researcher, and I did.
• 10:22
  SPEAKER_00: And sometimes I would consult with the biostatistician to help me with certain types of additional analysis that I didn't feel like I could do myself.
• 10:31
  SPEAKER_00: And some places like NIH or certain types of funding require it.
• 10:36
  SPEAKER_00: Okay?
• 10:36
  SPEAKER_00: And then, of course, the clinician, as we've been talking about, that perspective, which is the majority of us, right?
• 10:43
  SPEAKER_00: Should be all of us, right?
• 10:45
  SPEAKER_00: Even if you're a researcher, at some level, you're taking care of patients, right?
• 10:49
  SPEAKER_00: What was supposed to be done as you appraise this particular study?
• 10:54
  SPEAKER_00: And then, of course, there's the at the end of the study, you've completed it, you've collected the data, and now you're doing that summary statistics, the descriptive inferential statistics.
• 11:04
  SPEAKER_00: So I'm trying to keep us oriented into the timeline or the sequence of where we're going to be spending our time.
• 11:13
  SPEAKER_00: So I will always continue to reorient us back to that for that purpose.
• 11:18
  SPEAKER_00: Okay?
• 11:19
  SPEAKER_00: Now, why are we doing this?
• 11:21
  SPEAKER_00: Why does this all matter?
• 11:23
  SPEAKER_00: Because at the end of the day, you cannot, and nor can any researcher, study the entire population.
• 11:30
  SPEAKER_00: Uh even if it's a specific group within the population.
• 11:34
  SPEAKER_00: You just can't do it.
• 11:35
  SPEAKER_00: So you have to take a sample out of the population, okay?
• 11:42
  SPEAKER_00: You have to study them, and you maybe make an attempt to make it a judgment that is sound, that is internally valid, so that you can make inference back, hopefully with some degree of external validity and internal validity followed by external validity, so that the results that you're applying, you will see.
• 12:04
  SPEAKER_00: You should, they should manifest.
• 12:05
  SPEAKER_00: And if you think about the steps of evidence-based medicine that I lecture on in my EBM foundation series, step five is once you've applied the research and evidence to your patient care scenario, you're supposed to be assessing.
• 12:22
  SPEAKER_00: That's the fifth A and the five A's.
• 12:24
  SPEAKER_00: Assessing your own clinical practice and application of the research to discern are you getting what the results in the study also got?
• 12:37
  SPEAKER_00: Okay.
• 12:38
  SPEAKER_00: And so that's why the quality of research matters so much.
• 12:41
  SPEAKER_00: And you as a clinician need to be certain that what you're reading is internally valid.
• 12:48
  SPEAKER_00: And if you're the one doing the research, you want to work hard to develop a very well uh study design that is robust, that has integrity, that sustains itself through control of confounders and things like that that would affect the internal validity of the results.
• 13:12
  SPEAKER_00: Now, we're gonna be focusing a little bit of that time of our time today on getting you oriented to those two phases.
• 13:25
  SPEAKER_00: The pre-study phase, which is that statistical analysis plan or the SAP.
• 13:32
  SPEAKER_00: Some people also use it as the tick uh say like the statistical decision tree.
• 13:38
  SPEAKER_00: So it's like an algorithm that you walk yourself through.
• 13:41
  SPEAKER_00: And I'm gonna walk you through some of those questions to get you oriented to what that looks like.
• 13:48
  SPEAKER_00: So we're gonna cover the SAP process.
• 13:50
  SPEAKER_00: Okay?
• 13:51
  SPEAKER_00: Now there's no definitive SAP process for everybody.
• 13:54
  SPEAKER_00: This is the one that I've learned and used and all that, and there, and you'll see it published some in other variations of it, right?
• 14:02
  SPEAKER_00: But the idea is that you have a systematic approach to how you do research, even though different how you critically appraise a paper.
• 14:11
  SPEAKER_00: And then we're gonna move into the you close a study and the summary statistics, the descriptive and inferential statistics.
• 14:17
  SPEAKER_00: And so why do we do this?
• 14:18
  SPEAKER_00: Again, we want to make accurate interpretation of the results in its proper context so that we can make inference back to the population to decide.
• 14:32
  SPEAKER_00: And in the study, are we are we going to accept or reject the null hypothesis?
• 14:37
  SPEAKER_00: Remember, when you find no difference in a study, you have failed to reject the null.
• 14:42
  SPEAKER_00: Okay?
• 14:43
  SPEAKER_00: That means you're accepting it.
• 14:46
  SPEAKER_00: Said the same thing.
• 14:47
  SPEAKER_00: Okay.
• 14:47
  SPEAKER_00: You need to be able to look at it, especially for board exams.
• 14:50
  SPEAKER_00: I like to muddy the water the way I just wear that.
• 14:54
  SPEAKER_00: So if you fail to reject the null, you have accepted it.
• 14:58
  SPEAKER_00: You've not proven that it's wrong.
• 14:59
  SPEAKER_00: But you've got to make sure that you didn't make a type 2 error, right?
• 15:04
  SPEAKER_00: So where are we all going with this?
• 15:06
  SPEAKER_00: Well, I'm not gonna dive into this in detail right now.
• 15:10
  SPEAKER_00: And for those of you on the podcast who can't see this, I've got a table that is the kind of the typical standard statistical test that are usually done in a decision tree, okay, when we as clinicians are thinking about them.
• 15:25
  SPEAKER_00: Things like, you know, we're dealing with ordinal and nominal data, and then how many groups do we have?
• 15:31
  SPEAKER_00: Are they related?
• 15:31
  SPEAKER_00: Are they independent from each other?
• 15:33
  SPEAKER_00: And then that really guides us into which cell we land in and what kind of study or analysis, statistical analysis should have been done.
• 15:44
  SPEAKER_00: You know, we commonly hear about the students' t test or chi-square.
• 15:49
  SPEAKER_00: That's what this table is ultimately going to get you, you know, help you uh understand how to that if that was appropriate or not.
• 15:58
  SPEAKER_00: Now, listen, when we go through the SAP, this is done ahead of time.
• 16:02
  SPEAKER_00: So a good researcher maps this out very clearly in the methods for you.
• 16:08
  SPEAKER_00: Okay, they may not use that word SAP interpret uh literally, but they're giving you the pieces of the puzzle.
• 16:18
  SPEAKER_00: And so you need to be oriented, if you're developing your protocol, of this fundamental decision table as a piece of what you're going to do.
• 16:28
  SPEAKER_00: It's not the only thing.
• 16:30
  SPEAKER_00: Okay.
• 16:31
  SPEAKER_00: Most commonly, this is in the context of categorical data being the independent variable, the intervention that's being made.
• 16:39
  SPEAKER_00: Did you give them the drug or not give them the drug?
• 16:41
  SPEAKER_00: Okay, versus quantitative.
• 16:44
  SPEAKER_00: All right.
• 16:44
  SPEAKER_00: Now, so that's kind of one of the endpoints.
• 16:48
  SPEAKER_00: I show that to you so you know where we're going.
• 16:50
  SPEAKER_00: Okay?
• 16:51
  SPEAKER_00: So again, orientation, you want to know where you're starting and then where you're orienting to the end.
• 16:56
  SPEAKER_00: Some terminology that people get most, I think, bent out of shape about and confused is this concept of understanding the type of data variable the study was assessing, which is the primary endpoint.
• 17:13
  SPEAKER_00: Right?
• 17:13
  SPEAKER_00: So you have to define that.
• 17:15
  SPEAKER_00: That's your part of your hypothesis.
• 17:17
  SPEAKER_00: So again, I've done a lecture on what is a null, an alternative hypothesis, how do you create them, what do they mean, how do you try to protect yourself from making errors in either direction, type 1's and type 2 errors.
• 17:31
  SPEAKER_00: So you can you can refer back to that if you need that review.
• 17:35
  SPEAKER_00: Okay.
• 17:36
  SPEAKER_00: But you have to read the variable endpoint that that study is trying to assess.
• 17:43
  SPEAKER_00: Because that determines what type of data you are.
• 17:46
  SPEAKER_00: Is it categorical?
• 17:48
  SPEAKER_00: And within categorical, there are really two big types: those that have no order or rank, and those that do seem to have an order and a ranking with them.
• 17:58
  SPEAKER_00: Okay?
• 17:59
  SPEAKER_00: Those typically is nominal and ordinal data respectively, and they make up non-parametric statistical analysis that you would do.
• 18:07
  SPEAKER_00: And so it matters to understand: am I dealing with a yes or no, a dichotomous variable, or do I have a sense of ranking like mild to moderate, severe, like with pain, right?
• 18:20
  SPEAKER_00: Or am I dealing with quantitative and numerical assessments where I measure something that follows a continuous data variable pattern, commonly like vital signs, or a lab test, or those classic examples of that.
• 18:37
  SPEAKER_00: Now, scales of measure are important.
• 18:39
  SPEAKER_00: Again, we're still on that terminology, because again, they're helping you to determine what which descriptive statistics are valid for you to use, descriptive, being like frequency, mode, median, and mean, those kind of things, right?
• 18:59
  SPEAKER_00: If parametric or non-parametric statistical tests are appropriate, and if other methods of association or modeling are also best or should be considered.
• 19:12
  SPEAKER_00: Okay.
• 19:13
  SPEAKER_00: So that's why this matters that you are oriented around the type of data that you're dealing with.
• 19:20
  SPEAKER_00: Okay?
• 19:21
  SPEAKER_00: So let's go through that orientation and some of the questions that many of us who develop the research protocols go through, but you should be asking yourself as the reader.
• 19:34
  SPEAKER_00: Okay?
• 19:35
  SPEAKER_00: So I'm trying to make sure that you guys stay connected to the fact because some of you are gonna be clinicians and you don't care about what researchers do, you're not gonna conduct research, and this stuff kind of annoys you.
• 19:46
  SPEAKER_00: But the reality is you have to be able to assess and interpret that as you read the paper.
• 19:52
  SPEAKER_00: And it's very clear as you start to do this more often.
• 19:55
  SPEAKER_00: Okay?
• 19:55
  SPEAKER_00: So, what is the independent variable?
• 19:58
  SPEAKER_00: That is the intervention most commonly in our kinds of papers and research that we're reading.
• 20:05
  SPEAKER_00: Okay.
• 20:06
  SPEAKER_00: The intervention could be a non-pharmacologic intervention or a pharmacologic intervention in most cases, or exposure.
• 20:14
  SPEAKER_00: Okay.
• 20:15
  SPEAKER_00: Is it categorical?
• 20:17
  SPEAKER_00: Okay, so it's not something that you can measure.
• 20:20
  SPEAKER_00: It's that kind of one or the other, okay?
• 20:23
  SPEAKER_00: Dichotomous at some level, yes or no.
• 20:27
  SPEAKER_00: Typically uses the standard non-parametric and parametric, non-parametric statistical analysis.
• 20:34
  SPEAKER_00: That's a table that you can that I showed you, and it is available at thisiswhy.health for free as a download.
• 20:41
  SPEAKER_00: Okay, so if that's that's a great study tool.
• 20:44
  SPEAKER_00: Because if you can memorize that table for board exams, that'll do you a lot of good.
• 20:48
  SPEAKER_00: If you're writing a protocol, it'll do you a lot of good.
• 20:51
  SPEAKER_00: If you're reading papers and you're doing a critical appraisal, it'll do you a lot of good to make sure you are assessing.
• 20:56
  SPEAKER_00: Because we've all been that journal club where we're giving a journal club, and we're like, well, is it statistics correct?
• 21:02
  SPEAKER_00: So there's always somebody in the audience, isn't there, who's some biostatistician expert who's going to pick it apart.
• 21:09
  SPEAKER_00: Right?
• 21:10
  SPEAKER_00: Again, get oriented, you can get most of the way to the answer, as if it's appropriate or not.
• 21:16
  SPEAKER_00: And then there's the quantitative independent variable, and these are interventions that have a numerically measurable unit or have counts, things like maybe age, height, weight, okay.
• 21:31
  SPEAKER_00: Again, that typically doesn't apply a lot to what all the things that we typically do.
• 21:38
  SPEAKER_00: But when you're breaking down the independent variable, it becomes very important at the beginning, even though we follow most commonly categorical in our interventions.
• 21:47
  SPEAKER_00: Do they were exposed or they weren't exposed, right?
• 21:49
  SPEAKER_00: Yes, no.
• 21:51
  SPEAKER_00: And there's not a sense of ranking with that.
• 21:53
  SPEAKER_00: But with quantitative, this is where the uh models of association and techniques get applied, the logistic requirements.
• 22:00
  SPEAKER_00: Regressions, doing simple and multiple linear regression analysis become very important, especially if it's that continuous or quantitative independent variable.
• 22:12
  SPEAKER_00: Okay, the intervention.
• 22:15
  SPEAKER_00: What about the dependent variable?
• 22:17
  SPEAKER_00: That's the outcome, the primary endpoint or secondary endpoints of that particular study or paper.
• 22:24
  SPEAKER_00: Okay.
• 22:25
  SPEAKER_00: This is where most of the statistical test determination are really oriented.
• 22:31
  SPEAKER_00: Because remember, most of our interventions fall in the categorical bucket that I just went through.
• 22:37
  SPEAKER_00: That was question number one.
• 22:39
  SPEAKER_00: It's the highest level.
• 22:40
  SPEAKER_00: What is the intervention?
• 22:41
  SPEAKER_00: Yes or no, or is it measurable?
• 22:43
  SPEAKER_00: Number two is what am I measuring at the end of this study?
• 22:46
  SPEAKER_00: Are you dead or alive?
• 22:48
  SPEAKER_00: Right?
• 22:50
  SPEAKER_00: What was the mean change in blood pressure?
• 22:54
  SPEAKER_00: How much weight did you lose?
• 22:58
  SPEAKER_00: Those are endpoints that have a type, you have to treat that endpoint by a certain way.
• 23:04
  SPEAKER_00: And again, it fits in that table that I showed you a minute ago.
• 23:07
  SPEAKER_00: What row are you following into?
• 23:10
  SPEAKER_00: Okay?
• 23:11
  SPEAKER_00: So you hopefully you're starting to see that as you you you dive into this table, it points you to the answer very clearly.
• 23:18
  SPEAKER_00: That's why I developed that table and put it together, and the way I did is because it was a great study tool for me, but it helped me when I was doing my research.
• 23:26
  SPEAKER_00: But when you think about the statistical analysis that are being done for nominal and ordinal, there again, there's there's no ranking, there's no order.
• 23:35
  SPEAKER_00: You're not able to measure something.
• 23:37
  SPEAKER_00: It's that one or the other.
• 23:38
  SPEAKER_00: Okay?
• 23:39
  SPEAKER_00: Again, so you're most commonly going to choose a non-parametric statistical test.
• 23:45
  SPEAKER_00: Now there's a lecture on what is parametric, what is non-parametric.
• 23:49
  SPEAKER_00: I don't want to dive deep into it, but basically the data is not following a normal distribution when it's non-parametric.
• 23:56
  SPEAKER_00: Whereas ordinal and categorical or categorical, okay, which is also, you know, again, it's ordinal data is categorical, but it has a sense of order, a ranking, that pain.
• 24:06
  SPEAKER_00: Mild, moderate, severe.
• 24:09
  SPEAKER_00: But the magnitude of difference between those is not the same.
• 24:12
  SPEAKER_00: When you get to continuous or quantitative endpoints, right, that's the something you can measure.
• 24:20
  SPEAKER_00: It has a certain numerical value, right?
• 24:23
  SPEAKER_00: Blood pressure, temperature, labs, right?
• 24:26
  SPEAKER_00: Uh, and then you have to decide how many groups am I dealing with here?
• 24:31
  SPEAKER_00: And those all come together in guiding you as to what type of statistical tests that you're going to use.
• 24:38
  SPEAKER_00: Okay?
• 24:39
  SPEAKER_00: So when you start thinking about study groups, how many num how many samples or study groups am I going to do a comparison with?
• 24:48
  SPEAKER_00: One, two, three or more?
• 24:51
  SPEAKER_00: Are those groups the same patients or are they independent of each other?
• 24:57
  SPEAKER_00: So when you think of a typical randomized controlled trial, patients are randomly assigned to one or the other group and you follow them forward in time until the end of the study.
• 25:08
  SPEAKER_00: They never cross over in a typical randomized controlled trial.
• 25:14
  SPEAKER_00: The patients in those two groups are independent of each other.
• 25:17
  SPEAKER_00: They're not the same people.
• 25:20
  SPEAKER_00: But if I cross them over and I do a crossover study, a randomized crossover trial, then the patients become the same people in both groups, though placebo or you know baseline intervention, compare it to the comparative intervention.
• 25:37
  SPEAKER_00: And so when they're the same people, right, then you that that's good many ways.
• 25:42
  SPEAKER_00: There's some value to that.
• 25:44
  SPEAKER_00: There's some negatives to it, and I covered that in the study design lecture.
• 25:49
  SPEAKER_00: But you need to be able to orient it because that's what leads you back to the standard statistical test.
• 25:56
  SPEAKER_00: Now, the again, the assumption here for most of these tests is that we're dealing with most of the time categorical independent variables, the intervention.
• 26:05
  SPEAKER_00: You did it or you didn't.
• 26:07
  SPEAKER_00: You got exposed or you weren't exposed.
• 26:08
  SPEAKER_00: That's a yes or a no, right?
• 26:10
  SPEAKER_00: So we've already kind of assumed that.
• 26:13
  SPEAKER_00: But when you get to this point, so you see over here, you got in your column right here on the far left the type of data, the scale of measure.
• 26:23
  SPEAKER_00: Am I dealing with nominal data, ordinal data, endpoint, or continuous?
• 26:30
  SPEAKER_00: Puts you in the right row.
• 26:32
  SPEAKER_00: And then you have to decide, well, how many groups or samples do I have that I'm making a comparison with?
• 26:39
  SPEAKER_00: And are they related or independent or paired?
• 26:42
  SPEAKER_00: And I'm going to go through other lectures that break those down in more detail.
• 26:46
  SPEAKER_00: But the goal here is to get oriented to why you're learning those other topics.
• 26:51
  SPEAKER_00: Because you basically, let's say you have a nominal endpoint and you have, you know, paired or related groups.
• 26:59
  SPEAKER_00: Well, that's the statistical analysis that should have been done.
• 27:03
  SPEAKER_00: And so then you'll read the paper and say, Did they do that?
• 27:06
  SPEAKER_00: And maybe they'll tell you why they didn't.
• 27:08
  SPEAKER_00: Maybe they started out with that plan and something happened and they had to pivot.
• 27:12
  SPEAKER_00: And they're telling you, or they should tell you.
• 27:15
  SPEAKER_00: Right?
• 27:16
  SPEAKER_00: But that's how you use the table.
• 27:19
  SPEAKER_00: It guides you through the process of picking the right statistical test for the type of study design and methodology that you are either reading or, if you're the protocol development stage, what you're designing.
• 27:35
  SPEAKER_00: Okay?
• 27:36
  SPEAKER_00: So I hope that I hope that that's helpful because that's what guides us to determining the best way to analyze the data.
• 27:46
  SPEAKER_00: And we should do this in an a priori, preemptive manner before the study was started.
• 27:53
  SPEAKER_00: Good researchers, good protocols do that, and a good IRB will expect it.
• 27:59
  SPEAKER_00: I hope that you guys found this basic introduction helpful at getting you oriented.
• 28:06
  SPEAKER_00: I know there's probably still a lot of question marks.
• 28:09
  SPEAKER_00: Some of you might be a little frustrated, some of you might be disappointed.
• 28:14
  SPEAKER_00: I'm going to dive into some of these individual other topics and make sense of them more without trying to drag make the lecture too long.
• 28:23
  SPEAKER_00: But this orientation lecture is critical for your framework to build on the other components at a deeper level.
• 28:31
  SPEAKER_00: And so thank you for joining me.
• 28:33
  SPEAKER_00: Please give us a like if you're on YouTube.
• 28:35
  SPEAKER_00: We really appreciate the support.
• 28:37
  SPEAKER_00: And continue to watch.
• 28:39
  SPEAKER_00: If you're looking for a series, check out our playlist at YouTube, play at the podcast channels, and at thisiswhy.
• 28:46
  SPEAKER_00: And again, tables that we put out here are some of those are available for free.
• 28:51
  SPEAKER_00: Download at thisiswhy.health for you to study with, share.
• 28:57
  SPEAKER_00: You can use it in your lectures, give it to residents and students if they're needing that kind of stuff.
• 29:02
  SPEAKER_00: Those are available to help all of us so that we can continue to propagate good clinical research that impacts our patients for good.
• 29:10
  SPEAKER_00: See you next time.