SEM/HLM

[bcliff]

---

Members do not see this ad.

My program requires two semesters of statistics but is going to offer an optional SEM/HLM hybrid course this upcoming Fall, which would be my 3rd semester of graduate statistics. Any recommendations on whether or not this is a class worth taking? I'm not sure if it'll be practically useful in the short term given the kinds of projects my lab is working on, but I'm trying to think ahead for internship/postdoc, especially given my interest in neuropsych.

Any input is appreciated!

 

[PsyDr]

Depends on what you want to do with your career and your aptitude for maths. Consulting, research, program design, test construction, etc- go for it. Pure psychotherapy- no reason. Assessments- good for reading research, but not super necessary. Great at maths and it'll be an easy A- hell yes. Math is hard for you- probably not.

 

[AcronymAllergy]

I'm not a math whiz, but found the material to be interesting, and was able to do well in the class back when I took it (taught by one of our I/O profs). Same went for multivariate.

If you have the option and the time, I'd say go ahead. But that's based on my own situation in which I learn that type of material (or at least the basics of it) much better in a formal classroom-type setting than I do on my own.

 

[erg923]

My program requires two semesters of statistics but is going to offer an optional SEM/HLM hybrid course this upcoming Fall, which would be my 3rd semester of graduate statistics. Any recommendations on whether or not this is a class worth taking? I'm not sure if it'll be practically useful in the short term given the kinds of projects my lab is working on, but I'm trying to think ahead for internship/postdoc, especially given my interest in neuropsych.

Any input is appreciated!

Also not a mathematics whiz, but this knowledge helps one thinks critically about the subject matter. I think there has been a relative lack of practicing psychologists good at thinking methodologically and statistically for a log time. Again, I'm better I'm not the best, but think better trained than alot in the practicing community around here.

 

[Ollie123]

If you plan on long-term involvement in research - absolutely. They are widely used and pretty much necessary to have some familiarity with these days.

If you are looking for a purely practice-based career, I'd still consider it but I wouldn't push it as strongly.

I think I had 2 internship interviews where they cared that I knew things like HLM/GEE. Both were research powerhouses with lots of clinical trials going on and wanted me to be able to help out. At 95% of internships, it won't make much of a difference.

 

[bmedclinic]

My program required 3 stats courses, and this one came up and I took it a few years ago. Not only was it a blast, but it really helped me better think about SEM when reading articles. I'm pretty unlikely I think to ever publish a paper myself in that area, but I dig stats, and felt like it helped me understand the models ALOT better.

EDIT: That was 5 years ago now.

 

[AcronymAllergy]

My program required 3 stats courses, and this one came up and I took it a few years ago. Not only was it a blast, but it really helped me better think about SEM when reading articles. I'm pretty unlikely I think to ever publish a paper myself in that area, but I dig stats, and felt like it helped me understand the models ALOT better.

EDIT: That was 5 years ago now.

Agreed. And relatedly, it helped show me just how dangerously easy it is to setup and run analyses in Amos without actually having any idea what you're doing.

 

[smalltownpsych]

I say go for it. The more math, the better. Understanding the concepts underlying the statistical analyses just makes it so much easier to understand the research. When I was in undergrad and didn't even understand statistical significance and correlation coefficents, the research articles were like Greek to me. Now I am that way with SEM. I could really use that course. 🙂

 

[bmedclinic]

I say go for it. The more math, the better. Understanding the concepts underlying the statistical analyses just makes it so much easier to understand the research. When I was in undergrad and didn't even understand statistical significance and correlation coefficents, the research articles were like Greek to me. Now I am that way with SEM. I could really use that course. 🙂

Were it a different topic, I'd just say buy the book and do it self guided. However, I dont think such a rec is appropriate for SEM. You really want someone there to nicely tell you that your model is terrible and doesnt make sense until you get it.

 

[Pragma]

I took a couple of elective stats courses and they have paid off immensely. I think it is a worthwhile endeavor.

Only 2 required stats courses? That seems a little low to me.

 

[AcronymAllergy]

I took a couple of elective stats courses and they have paid off immensely. I think it is a worthwhile endeavor.

Only 2 required stats courses? That seems a little low to me.

I want to say my program actually only required two as well, in addition to tests/measurement and research design. I'd have to double-check that, though.

 

[Pragma]

I want to say my program actually only required two as well, in addition to tests/measurement and research design. I'd have to double-check that, though.

Yeah I suppose to some extent it matters how much you are going over statistics in related coursework, and also how much mentoring you are getting about doing statistics in your lab. However, I'm of the opinion that we should be pushing more statistics in graduate school, not less.

 

[AcronymAllergy]

Yeah I suppose to some extent it matters how much you are going over statistics in related coursework, and also how much mentoring you are getting about doing statistics in your lab. However, I'm of the opinion that we should be pushing more statistics in graduate school, not less.

Don't at all disagree. I would've liked to have had more course offerings (I ended up just with intermediate, multivariate, and SEM), but I think you're right in that they figured most students were getting more than enough additional tutelage during the required RA duties and in other classes.

 

[Justanothergrad]

Yeah I suppose to some extent it matters how much you are going over statistics in related coursework, and also how much mentoring you are getting about doing statistics in your lab. However, I'm of the opinion that we should be pushing more statistics in graduate school, not less.

I agree, although I don't know that more coursework is the way to go. I'm yet to find stats courses nearly as useful as self-study during a project. I find that it gives more room to understand the data when its something you design versus a dataset you are given.It lets you get more involved in the dataset and you work with it more over the course of a few months versus a week before moving on from a given analysis to a new dataset.

Don't at all disagree. I would've liked to have had more course offerings (I ended up just with intermediate, multivariate, and SEM), but I think you're right in that they figured most students were getting more than enough additional tutelage during the required RA duties and in other classes.

I don't know many (any really, but I'll assume I'm overlooking some) GRAs that get advanced stats training as part of that gig. I'm sure there are some, but I don't think thats most RAs by any stretch.

 

[AcronymAllergy]

I agree, although I don't know that more coursework is the way to go. I'm yet to find stats courses nearly as useful as self-study during a project. I find that it gives more room to understand the data when its something you design versus a dataset you are given.It lets you get more involved in the dataset and you work with it more over the course of a few months versus a week before moving on from a given analysis to a new dataset.


I don't know many (any really, but I'll assume I'm overlooking some) GRAs that get advanced stats training as part of that gig. I'm sure there are some, but I don't think thats most RAs by any stretch.

I was speaking specifically to my program, and why they might not have offered additional formal coursework. And the learning primarily occurred via the method you've mentioned--working on research and having the POI there to guide you in terms of new analyses.

Although I should also mention that my SEM course actually did require collection and analysis of unique data. Wasn't up to par in terms of being publishable, at least unless you went above and beyond the requirements (and happened to have an IRB in place already; and if that were the case, the instructor encouraged us to use it), but it definitely helped get a feel for working with "real" data. It was a particularly awesome prof who taught the course, though.

 

[Pragma]

I agree, although I don't know that more coursework is the way to go. I'm yet to find stats courses nearly as useful as self-study during a project. I find that it gives more room to understand the data when its something you design versus a dataset you are given.It lets you get more involved in the dataset and you work with it more over the course of a few months versus a week before moving on from a given analysis to a new dataset.

I think self-study is inadequate for really complicated statistics - maybe it works for you, but I don't think it would be sufficient for your modal graduate student.

The course value completely depends on the instructor. My stats professors were rock stars and kept everything applied/relevant. One of my grad school publications was a paper I wrote in a Multivariate course - the course intended for the outcome to be a paper you could publish. I felt that all of my stats coursework experiences were very worthwhile, particularly for things like SEM.

Mentoring statistics in labs was the norm in my program too. Not sure where a graduate assistant WOULDN'T get this unless they weren't working on research?

 

[Justanothergrad]

I think self-study is inadequate for really complicated statistics - maybe it works for you, but I don't think it would be sufficient for your modal graduate student.

The course value completely depends on the instructor. My stats professors were rock stars and kept everything applied/relevant. One of my grad school publications was a paper I wrote in a Multivariate course - the course intended for the outcome to be a paper you could publish. I felt that all of my stats coursework experiences were very worthwhile, particularly for things like SEM.

Mentoring statistics in labs was the norm in my program too. Not sure where a graduate assistant WOULDN'T get this unless they weren't working on research?

I've worked in several large grant funded RA labs and (speaking of folks I know who do the same in several highly ranked R1 universities), most RA work is not the type where one learns stats at all. Perhaps its different if the lab is within the program, but many RA slots are outside of psychology programs and those do not tend to involve that type of training (in my experience). As for statistics, I will say that not all professors are able to do the modern statistics and can offer very little guidance on many of these areas (I'm speaking of APA fellows, representatives, and editors of journals- not individuals who are far removed from the field/research).

I'm yet to find reading books and doing research projects with consultation to be insufficient to learn analyses. It worries me a bit if psychologists are unable to train themselves to conduct new research without taking a course. It worries me not just because researchers need to keep up with trends, but also because it makes me wonder if the same would be said of clinical skills. I think it may be harder, but that depends on the instructors. I've seen (and heard) of some very horrible statistics teachers.

I was speaking specifically to my program, and why they might not have offered additional formal coursework. And the learning primarily occurred via the method you've mentioned--working on research and having the POI there to guide you in terms of new analyses.

Although I should also mention that my SEM course actually did require collection and analysis of unique data. Wasn't up to par in terms of being publishable, at least unless you went above and beyond the requirements (and happened to have an IRB in place already; and if that were the case, the instructor encouraged us to use it), but it definitely helped get a feel for working with "real" data. It was a particularly awesome prof who taught the course, though.

I've seen one stats course that did that with data collection (it was a scale development course) and yes, you're entirely right, the outcome for that course seemed far better and the folks that went through it got more out of it. It makes me wonder why more classes don't take that approach.

 

[Therapist4Chnge]

I want to say my program actually only required two as well, in addition to tests/measurement and research design. I'd have to double-check that, though.

Same…but there were elective stats courses that were pretty popular. I wish I took the elective ones because on internship I felt a step behind sometimes, though I was able to read up and eventually feel on more even footing by fellowship.

 

[Pragma]

I've worked in several large grant funded RA labs and (speaking of folks I know who do the same in several highly ranked R1 universities), most RA work is not the type where one learns stats at all. Perhaps its different if the lab is within the program, but many RA slots are outside of psychology programs and those do not tend to involve that type of training (in my experience). As for statistics, I will say that not all professors are able to do the modern statistics and can offer very little guidance on many of these areas (I'm speaking of APA fellows, representatives, and editors of journals- not individuals who are far removed from the field/research).

I'm yet to find reading books and doing research projects with consultation to be insufficient to learn analyses. It worries me a bit if psychologists are unable to train themselves to conduct new research without taking a course. It worries me not just because researchers need to keep up with trends, but also because it makes me wonder if the same would be said of clinical skills. I think it may be harder, but that depends on the instructors. I've seen (and heard) of some very horrible statistics teachers.


I've seen one stats course that did that with data collection (it was a scale development course) and yes, you're entirely right, the outcome for that course seemed far better and the folks that went through it got more out of it. It makes me wonder why more classes don't take that approach.

Let me qualify my reply I suppose - I am thinking of doctoral student RAs, not hired RAs. Moreover, I am talking about graduate students, not psychologists. Of course self-study and consultation can be an appropriate method of learning for a psychologist. But I wouldn't replace core organized learning opportunities as a graduate student with just "self study" later on. There would be plenty of opportunity for misunderstanding. I notice some of this when I review manuscripts - sometimes it's clear that the authors have no idea what they are doing in their analyses.

I worked under some large grants as a student so maybe we had more resources, but we had statistics consultants that regularly worked with us on projects that we were doing (HLM, SEM, LGC, etc) with the datasets. We did a lot of hands on learning. Even myself in the supervisory role I took on as an advanced graduate student - a large portion of my time involved walking less experienced people through how to do different analyses and discussing what they meant. There was some self study - I remember having to do a really complicated power analyses for a grant proposal because we didn't want to bug the consultants until it was almost ready, so I was buried in books and articles at the library trying to figure it out. GPower wasn't going to cut it for this many variables and timepoints.

Of course there are also great stats workshops and talks that can be helpful as well.

Anyways, figuring out how to do statistics through mentoring was a major component of my doctoral experience. I believe it was one of the most important parts. So to just brush it off seems very odd to me, although I am intrigued if others aren't getting any mentoring in their research labs and resort to solely self-study.

 

[Justanothergrad]

Let me qualify my reply I suppose - I am thinking of doctoral student RAs, not hired RAs. Moreover, I am talking about graduate students, not psychologists. Of course self-study and consultation can be an appropriate method of learning for a psychologist. But I wouldn't replace core organized learning opportunities as a graduate student with just "self study" later on. There would be plenty of opportunity for misunderstanding. I notice some of this when I review manuscripts - sometimes it's clear that the authors have no idea what they are doing in their analyses.

I worked under some large grants as a student so maybe we had more resources, but we had statistics consultants that regularly worked with us on projects that we were doing (HLM, SEM, LGC, etc) with the datasets. We did a lot of hands on learning. Even myself in the supervisory role I took on as an advanced graduate student - a large portion of my time involved walking less experienced people through how to do different analyses and discussing what they meant. There was some self study - I remember having to do a really complicated power analyses for a grant proposal because we didn't want to bug the consultants until it was almost ready, so I was buried in books and articles at the library trying to figure it out. GPower wasn't going to cut it for this many variables and timepoints.

Of course there are also great stats workshops and talks that can be helpful as well.

Anyways, figuring out how to do statistics through mentoring was a major component of my doctoral experience. I believe it was one of the most important parts. So to just brush it off seems very odd to me, although I am intrigued if others aren't getting any mentoring in their research labs and resort to solely self-study.

Ah, yes. I think labs with functional research programs that produce consistent and thematic lines of research are able to do that sort of training. I thought you meant GRA positions. I think that good RA labs are less common than we give credit for and some of the labs that produce (not all, there are some awesomely amazing lab groups) well known lines still don't do this training. For instance, I've seen some of the labs I know, first hand, that don't do this with statistics training get mentioned on this board as examples/exemplars of great research training. It's something that has seemed odd to me as well. I do think that it is very important (stats and the mentoring that can come with it), but I don't think its a regular thing. I think courses and tutoring gives you a cursory and broad overview of an analysis whereas an application/self-study gives a very deep awareness of a much more narrow area. I've seen very few people come out of a statistics course and think "I've got everything I need to use this statistic in a publication". Perhaps this is my biased experience.

I don't separate advanced graduate students and psychologists in their ability to self-teach, in part because I am unsure why approach to learning would be different in the two groups. I would hope it wouldn't be and would be unsure why they were if, in fact, there were different.

 

[Pragma]

I don't separate advanced graduate students and psychologists in their ability to self-teach, in part because I am unsure why approach to learning would be different in the two groups. I would hope it wouldn't be and would be unsure why they were if, in fact, there were different.

From a practical standpoint you may be right, but I don't envision great outcomes for people that don't take stats coursework in graduate school and only go with self-study. There is still a component of "you don't know what you don't know" that I think gets hashed out (ideally) through the process of graduate school. There might be some really gifted students out there that can pick it up pretty easily, but I wouldn't trust their capabilities as a psychologist if there were not some metric for them demonstrating some minimal level of understanding. If you don't take the initiative to learn how to do basic statistics, regression, multivariate, factor, SEM, HLM, then how would you learn how to evaluate the quality of a paper that discusses things you might want to apply in your clinical practice?

My experience with stats courses again, was mostly applied and we used our own datasets. I don't think I would have become as strong of a researcher without that foundation. If the classes were crap then maybe I would feel different. I do disagree with your idea of not knowing enough from a course to publish something in an area - my experience was a bit of the opposite. Aside from the fact that clinical journals sometimes have crappy stats in their papers, it's not like the papers are also reporting on all of their assumptions, etc. I've just seen too much crap both published or as a reviewer to really have that much faith.

 

[Justanothergrad]

From a practical standpoint you may be right, but I don't envision great outcomes for people that don't take stats coursework in graduate school and only go with self-study. There is still a component of "you don't know what you don't know" that I think gets hashed out (ideally) through the process of graduate school. There might be some really gifted students out there that can pick it up pretty easily, but I wouldn't trust their capabilities as a psychologist if there were not some metric for them demonstrating some minimal level of understanding. If you don't take the initiative to learn how to do basic statistics, regression, multivariate, factor, SEM, HLM, then how would you learn how to evaluate the quality of a paper that discusses things you might want to apply in your clinical practice?

My experience with stats courses again, was mostly applied and we used our own datasets. I don't think I would have become as strong of a researcher without that foundation. If the classes were crap then maybe I would feel different. I do disagree with your idea of not knowing enough from a course to publish something in an area - my experience was a bit of the opposite. Aside from the fact that clinical journals sometimes have crappy stats in their papers, it's not like the papers are also reporting on all of their assumptions, etc. I've just seen too much crap both published or as a reviewer to really have that much faith.

Yeh, you're probably right. I think its more an intrinsic thing- folks who don't take stats (or take an interest in it, which is traditionally indicated by taking courses) do tend to have less skills at incorporating those things. I sometimes wonder what degree of 'lifelong learner' is taken with respect to ensuring that information is digested comprehensively versus by abstract (and by that I mean article abstracts as a source of all the info). It just seems like its an aspiration more than an embodied approach for many folks when it comes to an area that is not directly related. I see a lot of clinically focused folks hold back a gag reflex at the mere mention of research.

Your cynicism on research awareness is welcomed next to mine 😉

 

[Therapist4Chnge]

I worked under some large grants as a student so maybe we had more resources, but we had statistics consultants that regularly worked with us on projects that we were doing (HLM, SEM, LGC, etc) with the datasets. We did a lot of hands on learning.

Being able to know what/how to ask about a particular analysis has been a larger component of my collaborative research work than I initially anticipated. At the faculty level I have access to plenty of stats resources (both through the university and also outside stats consulting firm through a couple of diff large grants), but I feel like I'm regularly reading up on various approaches. This is a good thing because my knowledge would atrophy or become outdated otherwise.

Whenever people ask about, "should I take another stats course?" I almost always say yes because being in a classroom setting allows you to really dig into a topic and while there are opportunities to do that, "in the real world", it can be hard to do it well. While most clinicians won't regularly need/use SEM, LHM, etc…I think the process of learning it can help when reviewing journal articles on new clinical interventions/assessments.

 

[CheetahGirl]

I haven't read all the previous threads, but I agree with the perspective to take the course. I used SEM in my dissertation and there was a lot of self-study on my part. I would have loved to have taken a course in it prior to all of what I did. In the end, I have a very good understanding of how different ideas can inform different models, which leads to a sophisticated understanding of plausible clinical models (i.e., clinical constructs will surely inform your stats models once you know more about what is out there and how to use it). Andrew Hayes is my hero - fascinating person. I say 'dive in," it will make stats fun again (if you like math).

 

[Harry3990]

Would agree with everyone else that taking more stats is beneficial. Even if you don't actively contribute to research in the future, having that ability to quickly read through the methods/results of an article and automatically have a more thorough grasp of any limitations of the statistical methods being employed is very useful IMO. In my limited experience, some clinicians who do not have thorough statistical backgrounds tend to just trust what is being stated by the abstract or discussion. While you may hear the typical rationalization of "it was published in a peer-reviewed journal, so it must be valid," I prefer to encourage people to dig a little deeper and see whether you come to the same conclusion based on the evidence they are presenting. This is such a huge benefit of getting a PhD; it allows you to dive more into these intricacies and allows you to sharpen your critical understanding of the conclusions that are being inferred from statistical results in these papers.

 

[Pragma]

I haven't read all the previous threads, but I agree with the perspective to take the course. I used SEM in my dissertation and there was a lot of self-study on my part. I would have loved to have taken a course in it prior to all of what I did. In the end, I have a very good understanding of how different ideas can inform different models, which leads to a sophisticated understanding of plausible clinical models (i.e., clinical constructs will surely inform your stats models once you know more about what is out there and how to use it). Andrew Hayes is my hero - fascinating person. I say 'dive in," it will make stats fun again (if you like math).

Oh man, I used SEM in my dissertation too and my SEM course set me up really well for it. I think without that course, only with self-study - I might still be in grad school!