[Interview]: Interview with a cookie fanatic: Get to know Jake

[Jake]

1) How did you get into social psychology? What was your graduation?

I felt pretty certain that I wanted to get a PhD and end up as a university professor. As an undergraduate I believed, and still believe, that there were any number of fields I could have gone into and been pretty much perfectly happy. I ultimately decided to pursue social psychology because I decided it was the most interesting out of the options I was considering (which included cognitive psychology, evolutionary biology, and ecology).

2) What motivate you to a regular visitor in TS

1. Procrastination
2. The hope that I will learn about new statistical methods from others
3. The desire to stay sharp on statistical methods that I don't use often in my own research

3) Do you think human can beat bots? Does raptors are protector of human kind?

I think, given the chance, bots and raptors both would brutally subjugate humankind, and humans would not stand a chance. I for one welcome our new overlords whether they be robotic or reptillian.

4) Favourite food/drink?

Food: It's hard to beat a really good portabello mushroom burger...
Drink: Chocolate almond milk!

5) What are your hobbies?

I've been playing guitar for about 10 years and accompanying this with singing for about 5. I play at least a little bit almost every day and hate having to go more than a couple of days without access to my guitar.

Sadly I spend a lot of time reading dry academic papers/books, and after doing it for so long, I even kind of like it... =\

I love to play 8-ball and it is probably the only sport I've ever been any good at (if you do consider it a sport).

This song is for you (probably a punishment). VIDEO

Thanks

Jake, please tell us the top six to 10 characteristics about yourself you enjoy them or are proud of.

Then please name the worst six to 10 characteristics you hate about, or you are not proud of, or you are planning to change them in future (this one is 100% optional).

Oh god, the "strengths and weaknesses" question... I'm definitely going to pass on this one, sorry...

Don't you regret that you left programming as the main profession of your life? Don't you miss it?

I have a small amount of regret in leaving computer science. When I decided to leave that field, it was because I figured that my possible careers in that field basically included being a programmer, network engineer, etc. At the time the idea of going to graduate school for computer science was a completely foreign idea that just never occurred to me. Now I can see that if I had ended up going that route (the computer science PhD route) I probably would have been pretty happy with it. But I am also pretty happy where I did end up, so I don't really worry about it.

Which aspect of psychology do you like the most? (minimum three items!)
Why?

1. Well, first and most obviously, I just find psychological questions extremely interesting. A lot of the questions that psychologists deal with are questions that I have always spent a lot of time thinking about.
2. I like that there is still a lot of room for strong theorizing in psychology. I think the chances of one making an important theoretical contribution to a field like e.g. physics are pretty close to nil. But psychology is still young and undeveloped enough that there is a nontrivial chance one can actually have an influence on the field if you are good and/or lucky.
3. I also like that a lot of areas of psychology routinely deal with methodological issues that can be exceedingly complex if you let them be, so there are a lot of opportunities to apply advances in e.g. statistics to problems in psychological research.

Which aspect of it you don't like or you hate? (minimum 3!)
Why?

1. It is sometimes said that we are the Rodney Dangerfield of the sciences (we "get no respect").
2. "So are you, like, analyzing me right now?"
3. I think there are a lot of problems with the way we educate young psychology students. In a way there are 3 faces of psychology: (a) psychology as seen by the public, (b) psychology as seen from the perspective of an undergraduate psychology major, and (c) psychology as seen by psychologists. Unfortunately there are some rather huge gulfs between these 3 views of psychology. In particular a lot of undergrad psych majors find themselves unprepared and misled when it comes time to start doing actual psychology research. I will just leave it at that.

Jake, I am currently reading: Mixed Effects Models and Extensions in Ecology with R (2009), by Zuur, Ieno, Walker, Saveliev and Smith.

In it they introduce ML and REML by saying that in most text books, the explanation of the two methods is either too complex for a layman, or REML is simplified too much by saying it is a "mystical way to correct for degrees of freedom".

Do you agree with this? and how would you explan these methods to a layman without over simplifying it?

I actually have Zuur et al. and have read most (but not all) of it. I find their characterization to be pretty much right on and I find their level of explanation of the issue to be excellent.

As for how I would explain it to a layperson without oversimplifying. To be honest, I think I wouldn't. Let's face it, the difference is really quite a mathematical subtlety and it's not at all obvious why we should expect a layperson to be able to easily and accurately understand it after only a brief explanation by someone like me. When I try to explain these general issues (not necessarily ML/REML specifically) to students with at least a little statistical grounding, I usually go about it from the perspective mentioned above, appealing to the intuitions that (a) as MODEL gets biggers and bigger, ERROR is necessarily going to get correspondingly smaller and smaller, (b) but it is desirable for our estimate of ERROR not to depend on how many things we happened to have put in the MODEL, (c) therefore we want to apply a correction to ERROR to account for the size of the MODEL. I am of course oversimplifying, so like I said, in answer to your second question: I wouldn't.

[trinker]

Visualization is important to statistics. What type(s) of visualization(s) are your favorite(s)? If it's some more obscure type maybe you could throw an example up.

[Jake]

Trellis plots are indispensable when working with mixed models. Since I work with mixed models a lot these days, I would probably have to pick those as my top choice right now. I also quite like violin plots.

[vinux]

Thanks Jake.

1) Favourite programming language other than R?
2) favourite case study in social psychology(SP). what would you suggest some reading in SP for somebody good at statistics and a beginner in SP.
3) Have you gone outside US? Favourite place for your ideal vacation.

[Jake]

Favourite programming language other than R?

Python.

favourite case study in social psychology(SP).

Well, my favorite psychology paper in general is a review paper that was written by two social psychologists, although the topic only loosely falls inside the traditional boundaries of social psychology: LINK

what would you suggest some reading in SP for somebody good at statistics and a beginner in SP.

Definitely this book: LINK

Have you gone outside US? Favourite place for your ideal vacation.

Sadly no... I know, I need to get out more. Within the US, I really like the Rocky Mountains during the summer and southern California during the winter.

[bugman]

I also quite like violin plots.

I never quite "got" these.

But the point of this post is to say, "nicely said" in reply to my question about ML and REML.

[hlsmith]

What are your favorite cookies? Are animal crackers really crackers? Have you ever seen the movie the Hustler? Are you glad Colorado left the Big 12? Do you plan on staying in Colorado. Do you feel like you have super cardio powers when you go to lower altitudes? When do you tentatively think you will finish your PhD?

[trinker]

Are Fig Newtons really fruit and cake or are they actually classified as a cookie?

[Jake]

What are your favorite cookies?

Oatmeal chocolate chip!

Are animal crackers really crackers?

Don't be fooled -- they are cookies deep undercover.

Have you ever seen the movie the Hustler?

No. Is it any good?

Are you glad Colorado left the Big 12?

It's hard to think of something that I care less about.

Do you plan on staying in Colorado.

As much as I would love to stay in Colorado, the odds of this actually ending up being a tenable option for me are very low. I'm pretty much going to have to go wherever the jobs are.

Do you feel like you have super cardio powers when you go to lower altitudes?

I have to be honest, I have tested this several times, and I have never noticed any difference. Of course there has usually been some confounding factors so it's hard to tell for sure (e.g., most of my tests have been when I usually visit my family in Oklahoma for a little while over the summers, and it's hard to run with that kind of heat and humidity).

When do you tentatively think you will finish your PhD?

The original plan has always been to graduate Spring 2014 and I remain confident that I can make it out by then. But ask me again in a year .

Are Fig Newtons really fruit and cake or are they actually classified as a cookie?

This is a tough one, but I'm afraid I'm going to have to say that Fig Newtons are not cookies.

[trinker]

Thank you wise cookie master. The Fig Newton problem has perplexed me for years now.

[trinker]

Are there any methods in statistics that you think are downright wrong and wished they'd go away? If so why?

[bryangoodrich]

How long have you been using this data.table package? As far as I can tell, you're the resident expert now lol

[Jake]

Sorry guys I totally forgot that there were unanswered questions.

Are there any methods in statistics that you think are downright wrong and wished they'd go away? If so why?

There is never any reason, as far as I can see, to do a median split. Some argue that, although they agree that median splits should be avoided in the actual analysis of one's data, they can nevertheless be useful during the presentation and summary of results. I think this view is also wrong. If you want to summarize what happens at the high and low end of some continuous predictor, you can just as easily talk about what happens with the predictor at, say, 1 standard deviation above and below the mean of that variable according to your model. In terms of requisite statistical knowledge this is no more demanding than understanding a median split, and it has the very desirable property that the numbers you report are derived from your actual statistical model, instead of being based on something that is conceptually similar but not really the same.

Also... one-tailed hypothesis tests. In practice they end up just being a somewhat sneaky way to double your alpha level. If this is what you want to do, then fine, but just say you are adopting a more liberal alpha level, don't feed me a dubious line about the observed effect being in the direction you privately predicted...

How long have you been using this data.table package? As far as I can tell, you're the resident expert now lol

I believe I started getting acquainted with it very shortly after joining this forum... so that would be Summer 2011.

[hlsmith]

Figs definitely aren't cookies. Would you want them with milk or ice cream or in a blizzard, not me. I lived in southern US for awhile and saw that chocolate chip cookies typically went flat, and figured thats why they have pralines. Are there any cookies that are better/worse in the altitude or is that why everyone in CO seem to be eating granola?

[Dason]

Just so it's documented somewhere other than the chatbox...

What are you thoughts on no bake cookies?