Steve Tippins explains how doctoral students should choose statistical tests by matching their research question, data type, and sample size. Non-significant findings are acceptable if you report limitations and suggest future research. Design your data collection around your intended analysis from the start.
Once you have completed interviews, transcribed interviews, then you're gonna come up with codes and themes and ultimate answers and finish degrees. Well, you know, good luck, and if we can help you in any way, reach out to us. So, how do I get the right statistical test? Again, based on your research questions. Is there causation? Am I talking about a relationship? Also, what type of data do I have? What do I do with missing data? Or if I have missing data? First, If I've got— I need that 142 people and I need, you know, 10 people haven't provided information, many times I just exclude them and I get 9 more participants or 10 more participants. Sometimes I'm going to say, okay, I've got 142 respondents, but on this question I've only got 136 respondents. So sometimes you can do it that way. Some people have said you can insert the mean of the data from those certain questions to put in. I don't think that that's as good an answer as possible, but typically it's collect more data is the ultimate answer. And I have remembered that piece of software, it's LAIRD, L-A-I-R-D Statistics. And if you use LAIRD, they will walk you through the steps that will tell you which is the appropriate test to use based on the data you're collecting. My chair wants more rigor. What does that actually mean? When, uh, I mean, that's a good question, and I, you know, I really don't like when they just kind of disguise what they're looking for. Oh, give me more. What that means typically is you're doing doctoral-level research. You have to do, in a quantitative study, you have to do relational or regression or something like that, tests that require more data and more involvement. A t-test, just comparing things, they're typically not going to accept. So they want to know. Now, it's up to them to tell you this, and it's up to them to approve what you're doing at the beginning as part of your proposal. So if they're asking you at the beginning, great. If they're asking you at the end where they've already approved what to do, a lot of the burden is on them too, because they let you move forward. So you say, you know, engage your chair in a conversation. Okay, by rigor, do you mean something like a regression test or ANOVA or something like And they may say yes, and then you start wording things from there. And a great place to go is previous dissertations that your chair has served on and similar types of research so they can— you can know what types of things they've approved in the past. Ah, what a great question! Oh my gosh, I've done all this work and my results are not significant. And the answer is, it really doesn't matter in a dissertation if your results are significant or not. To go to a journal, yeah, they want significant stuff, but the idea and goal of a dissertation is to show and demonstrate you can do research and that there's support for doing this research and support for your research questions. If you came up with non-significance, you report non-significance. You then say in your limitations and, um, Chapter 5 and future research, you say, "Oh, future research should be done too, and maybe we should be looking at different, um, groups or use different tests or whatever," but you get to report what you found. And if you found something was not significant, you report that. A lot of times you'll have multiple things you're testing, and maybe 3 of them are significant and 2 aren't. So you report the ones that are significant, you report the ones that aren't. That's just what research does. Sometimes it doesn't work, and that's good information to know. So reliability and validity are all based on setting things up properly at the beginning. After the fact, you've already collected it. Um, am I using the right type of people? Am I using the right test? Am I gathering the information in the appropriate manner? That's all gonna talk to— I'm just gonna use credibility for both of them. It's all gonna make your research credible and valid. If I'm using tests that haven't ever been done before or collecting data in a manner that's never been done before, it's harder, much harder. But if I'm following standard procedures, it's a lot easier to justify those areas. Your overall methodology and then which approach you take, the typical answer is, well, it's based on your research questions, based on the problem. Okay. In my mind, a lot of qualitative research is you're going to interview people. So it's what's the emphasis you're interviewing them based on. And if I'm looking at, for example, gosh, what's their experience in their life on this? Well, that's a lived experience and that's going to be a phenomenological approach. Or if I've got looking at respondents all from one group, that might be a case study. So, it should lead quite logically and naturally from what you're asking. Every school has, that I know of now, has some sort of Institutional Review Board, IRB. And they all have classes of people that are protected. That doesn't mean you can't do research on them, but that more and more precautions need to be, uh, put in place. Members of the military, prisoners, pregnant women, minors are typical groups, people, of people that are, um, considered protected. The other thing you have to be aware of when you're dealing with healthcare health issues is HIPAA, and it's issues like that. So if you can get access to patients without violating HIPAA and without, you know, protected classes, whether it's you've set up safeguards for them that your school agrees with, or you can do that in other times. There are people say, "Oh, I want to know what providers perceive pregnant women," in this example, "got from this level of intervention. So you're not asking pregnant women, you're asking people who've worked with pregnant women what their perceptions are. So that can be ways around things. Definitely thinking ahead of time is always good. Great. And understand also that you can ask me any question at all. I'm on your side. So your data method, data collection is based upon your research questions. So design those Well, think about analysis, how you're gonna analyze stuff while you're designing the data you're gonna collect and what data you need. That's gonna save you lots of time and headaches in the future. Everybody have a beautiful day. Bye-bye. Bye-bye.
