Those with Statistics Skills . . . your opinions please!

[sste]

So, in my work I am starting to more empirical, data-gathering types of projects. My statistics ability is extremely rudimentary and so I have either done simple things (think t-test, anova) or I have worked with a statistical consultant. So, these are my questions:

1. How long would it take studying on a part-time basis to acquire "very strong" statistical skills? In other words, I am wondering if I should make this investment - - if it is going to take 2-3 years for example, I am wondering if it is best to just continue to use consultants - - especially given that statistics change over time.

2. Is it helpful to learn/relearn matrix algebra and calculus? One intensive five week advanced statistics course (regressions, I think some SEM) I am looking into actually has a large component that is review of matrix algebra and calculus. Is this helpful in your opinion? For me it would not be review at this point but learn.

3. Can a so-so math person do statistics well? So, I did well in my college statistics course but it was work for me and I have never felt that math was a particular strength.

Thanks in advance for any opinions!

[egoldber]

You cannot do any advanced statistics work without knowing matrix algebra and advanced calculus well. If those are not skill sets for you, then that would be the first step. Otherwise it is conceptually difficult to study more than simple linear regression and very basic analysis of variance.

If you want to do advanced survey analysis or survey design work, you are looking at a fair amount of advanced statistical coursework.


Can a so-so math person do statistics well?

I guess it depends on what you mean by "so so" and "well". :)

If you found basic calculus relatively easy to learn, then I think so. If by "well" you mean able to apply basic concepts (including logistic regression, multivariate analysis of variance) to study situations and (more importantly) to know what you DON'T know and when to get help, again I think you are OK.

If you don't have any interest in taking things to that depth, then I think you are better off using consultants.

[sste]

Thank you! I think my goal is not to become an expert statistician but rather a very competent methodologist who, as you note, knows what I don't know and when to get help.

I don't have any memory whatsoever of matrix algebra and calculus which I took in high school except that I completely missed the boat in terms of keeping up at the beginning of calculus and that caused later problems! Sounds like I will need to start there . . .

For math ability, I would say I am probably good or very good relative to the general population but "so-so" relative to the people I encountered in my competitive college or even more so in my job now within academia. I am logical and actually always scored quite well on math type skills in IQ testing . . . but I have never been stellar at actual math.

[cvanbrunt]

Thank you! I think my goal is not to become an expert statistician but rather a very competent methodologist who, as you note, knows what I don't know and when to get help.

Beth is the stats expert around here and I would listen to her. I was going to post but you beat me to it. I'll say it anyway. I'd be more concerned with methodology. The best statistician can't fix poorly collected data.

[sste]

Carrie, are there classes or books or other things you would rec. for methodology? I am doing mostly psych and sociology work as it intersects with law (my primary field). I have done some empirical projects and I think have a decent knack for methodology, largely because I am a compulsive worrier who naturally anticipates every problem in life. I also took one class in advanced methodology recently that consisted of critiquing experiments and this was actually very helpful.

Should I take a class in survey design? Or something else? Or just try for some on the job training through co-authoring?

[egoldber]

my goal is not to become an expert statistician but rather a very competent methodologist

I hate to give the depends answer, but I think it depends. :p

If you want to actually contruct your own surveys, sample designs and not just do data analysis, then I think you really need to do more advanced coursework. If you want to analyze data that someone else designed for you, then perhaps not.

What is your field? I think it is hard in some fields to be the latter without also being, to some degree, the former. I am thinking in particular (because it is what I know :) ) about people in public health and epidemiology. The best researchers I know are also pretty darn competent statisticians even if that is not their degree field.

[tarahsolazy]

Depends on what type of studies you'll be developing. My research area is clinical trials and epidemiology, and I have an MPH (in addition to my MD) that taught me how to design studies. If you want to do survey studies, take some survey design. If you plan to do cohort studies, collect time series data, study interventions, etc, there may be epidemiology study design stuff that would be helpful.

I do my own statistics and have had to teach myself SAS programming, with the help of their LiveWeb classes. I don't know matrix algebra or advanced calculus, but I know enough about which modeling techniques to do with what kinds of data I collect and what types of analyses I want to do, ie logistic vs. linear or Poisson regression, survival, etc.

[justlearning]

I can't give you good advice about what you can/can't do or would/wouldn't learn in a class like that.

But I will say that I have never taken a calculus course (went to a bad high school) and only took one basic algebra course in college ("college algebra"). I was good at math but they would have only counted as electives towards my undergraduate degree so I didn't take any more math (nor did I in my grad program).

But as part of my graduate program I took advanced statistics courses and was always the top-performer in them. I even took two classes out of the math department and drove my friend (a PhD math student) crazy because I got better grades in those classes than he did! Why? Because I'm very analytical, can figure out how to apply my knowledge to the problem at hand, and can figure out how to accurately use computer programs to analyze data. I'm also detail-oriented so I'm always double-checking my work along the way. That skill set served me very well and I didn't really need to understand all of the mathematical concepts behind everything as long as I could understand conceptually what each test did.

Now I didn't take really advanced courses (like Beth would have) so I'm guessing that I went as far as I could with my limited math education. But I did learn how to do logisitc regression, MANOVAs, SEM, survival analysis, and other similar stats. (But after being a stay-at-home mom for so long now, I'd probably be struggling to just do a simple regression at the moment!)

So I think it is possible to do some advanced stats without having math education. But I do think that you need good analytical skills.

[sste]

So, I am in legal academia and the field is very unusual in not requiring a ph.d. Training is a j.d. which is not empirical/methodological. As a result, most of the research in law tends to be more normative - - x law should be changed in y way based on a, b, and c concerns. Increasingly there is interest in more empirical study in law. Since my background as a college student was in neuropsych, I have long felt it was pretty lame for law to be an academic discipline devoted to making unsubstantiated claims.

The things I would like to study empirically in the next decade or so include things like the effects of social networks on various legal behaviors and defaults, the effect of social capital on legal compliance, analysis of court cases to see if x or y is actually being done, and perhaps some public surveys of attitudes/opinions toward the legal system.

[cvanbrunt]

It really depends on the kind of work you are doing. If you are doing survey work, I'd look into a survey design course. Survey design is a specific field and it is very easy to do a crappy job.

I'm an experimental psychologist. I am biased, admittedly. But if you are going to take a general design class, take in a psych department. Psychology is the unwanted step-child of the natural sciences. For some reason, chemists and biologists have a hard time accepting that people are part of the natural world. (But that is an esoteric profession bitch on my part) As result, psychology has the most rigorous methodological requirements when designing studies.

Here are two great books. They are both classics. The second is an update of the classic by Cook and Campbell. Shadish taught me methods in grad school and I TAd for him.

http://www.amazon.com/Designing-Experiments-Analyzing-Data-Perspective/dp/0805837183/ref=sr_1_1?ie=UTF8&s=books&qid=1257525565&sr=1-1

http://www.amazon.com/Experimental-Quasi-Experimental-Designs-Generalized-Inference/dp/0395615569/ref=sr_1_1?ie=UTF8&s=books&qid=1257525622&sr=1-1

These are really for quantitative data. If you are looking at qualitative data, then that is another ball of wax. Someone who works in that realm would have other suggestions.

[egoldber]

Hmm. Your work sounds like an interesting mix of qualitative and quantitative.

I would agree that you don't need as much math for that (I am biased in thinking that the math is important :loveeyes: ), but a good course or two in survey design would be important.

[Nechums]

I think everyone answered wisely so far, so I don't have much to add other than to say how happy it makes me to see a thread on stats!

[cvanbrunt]

I just wanted to add another link to a book. I don't know the book but it's really to clue you into a series from Sage University Press. These are short little texts on very specific topics in stats and methods. I find them very useful.

http://www.amazon.com/dp/0803927436/ref=rdr_ext_sb_ti_sims_2

[egoldber]

The Sage publications tend to be very good. There are several really good survey ones. I think Kalton's Introduction to Survey Sampling is particularly good.

http://www.amazon.com/Introduction-Sampling-Quantitative-Applications-Sciences/dp/0803921268/ref=pd_sim_b_2

[larig]

My field is education research. Many, many researchers in my field who do quantitative research have no idea how to deal with the the underlying mathematical concepts they are using when they are analyzing their data using stat software. Those of us who are going get PhDs and do research in the field really don't have to take that many methods courses by requirement--I was surprised. Like others said, though, that just tells me that how much you need to know would depend on what kind of field you're in and what you're doing with the data, if you're collecting it, etc. I'd guess if you could just pick and choose the classes you were interested in (i.e., not pursue a degree and take classes at large) it wouldn't take you long to get that level of competency. Almost every field has its own versions of quantitative methods classes, so you should be able to find something that would work and be relevant.

[cvanbrunt]

Beth,
Have you ever taught undergraduate stats? I'm guessing you would have taught it in the math department and not psych but I thought I'd throw it out there.

My chairman emailed today and asked if I would be interested in teaching stats instead of methods. I love teaching methods but it is draining. It's a writing intensive course with a lab and it nearly kills me every time. Anyhow, I think I would like to give stats a shot. Do you know of any cool resources for stats teachers. I could search but today is a lazy day.....

Anyone out there teaching stats, feel free to chime in.

[larig]

The things I would like to study empirically in the next decade or so include things like the effects of social networks on various legal behaviors and defaults, the effect of social capital on legal compliance, analysis of court cases to see if x or y is actually being done, and perhaps some public surveys of attitudes/opinions toward the legal system.

I saw this after I posted my reply, but this stuff sounds like you'd find some great methodological classes in a communications department. (that is my outside of college methods concentration, myself, because of my interest in social networking & education). I took a class on internet research methodology from the comm dept. in my coursework--it was great, and consequently I asked the prof to be on my dissertation committee.

[sste]

You guys are incredible! Thank you so much. All of these books have been added to my amazon cart. I think I am going to look for a statistics grad student tutor to at least get me up to the simple regression point and then tackle calculus if I want to go further. And I am going to take a survey methods class in the spring as a start.

I really appreciate the advice.

[egoldber]

Have you ever taught undergraduate stats? I'm guessing you would have taught it in the math department and not psych but I thought I'd throw it out there.

I did as a grad student in stats and then for extra money occasionally before I had kids. It's very time consuming initially, but not so bad after that.

The thing I hated was the students who did not want to be there and by the nature of what I taught, there were a lot of those....