Where do experts go to learn more about statistics (in social science)?
January 12, 2016 11:39 AM Subscribe
If you do quantitative research for a living, where do you go when you have questions, or want to learn more about how to analyze data?
I'm getting ready to graduate from a social sciences PhD program and I do a lot of quantitative research. Even though I've worked on a lot of research projects, there's still a lot that I don't know. For instance, I've done a lot of work with regression and experimental designs, but almost nothing with time series and hierarchical linear modeling outside of some very cursory in-class work.
So: Where do you go to learn more about statistics? Can you recommend some resources that have taken you from basically zero knowledge to being able to do analysis for something like a peer-reviewed journal?
The obvious answer is to ask a colleague, which I do, but the number of people who I know who can deal with nuanced statistical problems and who would be willing to walk me through an analysis step by step is probably going to dwindle once I am no longer a student.
I'm getting ready to graduate from a social sciences PhD program and I do a lot of quantitative research. Even though I've worked on a lot of research projects, there's still a lot that I don't know. For instance, I've done a lot of work with regression and experimental designs, but almost nothing with time series and hierarchical linear modeling outside of some very cursory in-class work.
So: Where do you go to learn more about statistics? Can you recommend some resources that have taken you from basically zero knowledge to being able to do analysis for something like a peer-reviewed journal?
The obvious answer is to ask a colleague, which I do, but the number of people who I know who can deal with nuanced statistical problems and who would be willing to walk me through an analysis step by step is probably going to dwindle once I am no longer a student.
I'm a business school professor. The colleagues around me learn new research methods through one of the following methods:
1) Sit in an advanced statistics class being offered in a different department or nearby university, offering the techniques they are interested in.
2) Attend research methods workshops organized in conjunction with major conferences in our field.
3) Find a few other like minded individuals, and organize a workshop session at our college, bringing in known experts in the field (we had one on hierarchical linear modeling last semester for example, bringing in the authors of a popular software solution in the field).
4) And finally, buy books covering those methods and software; which are geared towards serious researchers (not the dummies books, but the academic press books).
posted by tuxster at 12:38 PM on January 12, 2016 [3 favorites]
1) Sit in an advanced statistics class being offered in a different department or nearby university, offering the techniques they are interested in.
2) Attend research methods workshops organized in conjunction with major conferences in our field.
3) Find a few other like minded individuals, and organize a workshop session at our college, bringing in known experts in the field (we had one on hierarchical linear modeling last semester for example, bringing in the authors of a popular software solution in the field).
4) And finally, buy books covering those methods and software; which are geared towards serious researchers (not the dummies books, but the academic press books).
posted by tuxster at 12:38 PM on January 12, 2016 [3 favorites]
Try the Tableau users forums - real big on meaningful data application and interpretation. They also have local user groups so you can get together with others in your field and interest.
posted by ptm at 1:31 PM on January 12, 2016
posted by ptm at 1:31 PM on January 12, 2016
What I did:
1. My university runs various free one day stats courses for grad students on lots of different topics. I attended those.
2. I audited some taught postgrad seminars for things that weren't covered.
3. Check what's online - Duke and Harvard seem to put LOADS of their coursenotes online, our lecturers used to link to them in their coursenotes.
4. As a post-doc, collaborate with somebody who does have these skills, and get them to teach you. This is the main reason collaborations exist in my experience (one person wants access to skills or resources that the other person already has). So in your case work with a statistician. They may have an interest in the problem you are looking to research, but have no experience in running a study. Both of you benefit.
5. If all else fails, pay to go on a course (distance or IRL). My institution gave us a study budget for that sort of thing so I could claim the money back.
posted by tinkletown at 1:32 PM on January 12, 2016
1. My university runs various free one day stats courses for grad students on lots of different topics. I attended those.
2. I audited some taught postgrad seminars for things that weren't covered.
3. Check what's online - Duke and Harvard seem to put LOADS of their coursenotes online, our lecturers used to link to them in their coursenotes.
4. As a post-doc, collaborate with somebody who does have these skills, and get them to teach you. This is the main reason collaborations exist in my experience (one person wants access to skills or resources that the other person already has). So in your case work with a statistician. They may have an interest in the problem you are looking to research, but have no experience in running a study. Both of you benefit.
5. If all else fails, pay to go on a course (distance or IRL). My institution gave us a study budget for that sort of thing so I could claim the money back.
posted by tinkletown at 1:32 PM on January 12, 2016
Good advice so far; I especially favor tuxster's 2, 3, and 4 (and look for books that emphasize application over theory). I'd add:
- Rather than looking for any old empirical article, look for one that provides the entire analysis script (like this one). They're still scarce but increasingly common.
- UCLA's IDRE group has a wide variety of pretty pragmatic tutorials, usually in more than one statistical language
- stats.stackexchange.com is a good supplemental resource when you have specific questions about a method.
Specifically for HLMs: I really, really like Gelman and Hill's book.
posted by McBearclaw at 1:32 PM on January 12, 2016 [1 favorite]
- Rather than looking for any old empirical article, look for one that provides the entire analysis script (like this one). They're still scarce but increasingly common.
- UCLA's IDRE group has a wide variety of pretty pragmatic tutorials, usually in more than one statistical language
- stats.stackexchange.com is a good supplemental resource when you have specific questions about a method.
Specifically for HLMs: I really, really like Gelman and Hill's book.
posted by McBearclaw at 1:32 PM on January 12, 2016 [1 favorite]
I have learned a lot by listening to audio recordings of specific courses, during my commutes and doing chores. A favorite, that may be of interest to you:
Anything by Lesa Hoffman, who provides videos, syllabi, sample code linked from here. It looks like the materials she links to are now all video, but it used to just be MP3 (preferable for me). She is a quantitative psychologist who specializes in multilevel (aka hierarchical) models for longitudinal/repeated measures data; but I think one of her courses talks about "classic" HLMs (e.g., people nested within schools/neighborhoods/sites). She also has/had a "multivariate" (i.e., latent variables, factor analysis, SEM) course which helped fill in many gaps that my in-person SEM course did not cover. She has a great sense of humor, and explains things in enough detail to understand and apply these complicated techniques, without it becoming a pure-stats course.
posted by mean square error at 1:39 PM on January 12, 2016 [5 favorites]
Anything by Lesa Hoffman, who provides videos, syllabi, sample code linked from here. It looks like the materials she links to are now all video, but it used to just be MP3 (preferable for me). She is a quantitative psychologist who specializes in multilevel (aka hierarchical) models for longitudinal/repeated measures data; but I think one of her courses talks about "classic" HLMs (e.g., people nested within schools/neighborhoods/sites). She also has/had a "multivariate" (i.e., latent variables, factor analysis, SEM) course which helped fill in many gaps that my in-person SEM course did not cover. She has a great sense of humor, and explains things in enough detail to understand and apply these complicated techniques, without it becoming a pure-stats course.
posted by mean square error at 1:39 PM on January 12, 2016 [5 favorites]
A friend of mine who has a PhD in a social science discipline and does tons of quantitative research speaks highly of the Sage book series "Quantitative Applications in the Social Sciences (QASS)". Each book addresses a specific technique and the series currently contains a total of 175 books. (Direct PDF download of a list of book titles in series).
For time series, see volume 9.
posted by obscure simpsons reference at 6:23 PM on January 12, 2016 [2 favorites]
For time series, see volume 9.
posted by obscure simpsons reference at 6:23 PM on January 12, 2016 [2 favorites]
Another strategy that I've seen be successful for many colleagues is the workshops that are offered as part of national/international conferences.
posted by zachxman at 10:37 AM on January 14, 2016
posted by zachxman at 10:37 AM on January 14, 2016
« Older Winter boots that actually grip | Toddler books with lots of things in the pictures? Newer »
This thread is closed to new comments.
- read a bunch of empirical articles that use it
- talk to a friend that has used it
- if I can swing it and I *really* need to learn it, I will audit a class at my university (I'm a professor) and/or take an ICPSR course
posted by k8t at 11:49 AM on January 12, 2016 [3 favorites]