Comments for [citation needed]

Comment on what do you get when you put 1,000 psychologists together in one journal? by guy incognito

guy incognito — Thu, 16 May 2013 08:25:31 +0000

awesome. is the code to make the interactive graph available somewhere?

Comment on R, the master troll of statistical languages by Wendell

Wendell — Tue, 07 May 2013 20:00:46 +0000

Thom: “even R fanatics don’t tend to use or advocate R for general programming – just for statistics.”

Not disputing your general argument, but I just finished reading the book, “Quantitative Corpus Linguistics with R”, in which R is promoted for text processing. I think it was the most mind-numbing misapplication of a programming language I have ever seen.

Comment on R, the master troll of statistical languages by Thom

Thom — Mon, 29 Apr 2013 09:21:33 +0000

John P. : “However, I can only see R fanatics insisting that this is a full tool all by itself because most apps in the world need statistics as ONE of the outputs of a piece of software.”

That’s a bit of a straw man – even R fanatics don’t tend to use or advocate R for general programming – just for statistics. Most R fanatics (that’s I’m aware of) will happily use other languages to call R (e.g., Python) or use R to manage other software (JAGs etc.).

Mind you, lots of R users probably over-use R in the sense that some other language would be more efficient for their task, but that’s because of switching and other costs. However, that’s a general rule of programming (or indeed technology).

Comment on R, the master troll of statistical languages by John P

John P — Mon, 29 Apr 2013 08:51:48 +0000

R gives the world its ten-thousand-and-first computer language. However, I have found that using R as a standalone language is a bad idea. It’s much, much better to prepare data for R, and to receive data from R, from a scripting language like Perl or Python or Ruby. The extraordinarily limited number of data types, the lack of pointers (references), and a host of other things make this tough sledding for people who are used to languages that can stand on their own.

R’s convenience functions for textual data are hilariously underpowered. It’s nice that R circumvents the bloatedness of old SPSS or SAS programs and it’s also nice that R is so easy to call from all the major scripting languages. However, I can only see R fanatics insisting that this is a full tool all by itself because most apps in the world need statistics as ONE of the outputs of a piece of software. That’s why your, for example, Perl script calls R as a kind of Perl convenience function … and it is VERY convenient for that.

Comment on the truth is not optional: five bad reasons (and one mediocre one) for defending the status quo by Improving Interpretation of Science Writing | Houses of Stones

Improving Interpretation of Science Writing | Houses of Stones — Sat, 27 Apr 2013 20:39:55 +0000

[...] scientists are blamed, and rightly so. Sometimes criticisms identify more systemic problems like current scientific practice, or scientific institutions like the NSF or a university, or academia in general. Blame is also [...]

Comment on Feynman’s first principle: on the virtue of changing one’s mind by Feynman’s first principle: on the virtue of changing one’s mind | SimoleonSense

Wed, 24 Apr 2013 17:06:07 +0000

[...] Click Here To Read: Feynman’s first principle: on the virtue of changing one’s mind [...]

Comment on what do you get when you put 1,000 psychologists together in one journal? by Kristina

Kristina — Wed, 10 Apr 2013 21:17:55 +0000

http://pilab.colorado.edu/assets/authordata.json is the data. The nods part is a list of the authors, the links is a list of how theyre related, using the indexes of nodes as the id for source and target.

Comment on what do you get when you put 1,000 psychologists together in one journal? by Gilad

Gilad — Sat, 06 Apr 2013 12:27:48 +0000

That’s really cool

Is that data available in its raw form?

Comment on the truth is not optional: five bad reasons (and one mediocre one) for defending the status quo by Tal Yarkoni

Tal Yarkoni — Fri, 15 Mar 2013 01:09:16 +0000

Rich, if you read this blog regularly you’ll know that I’m the last person who’d endorse slavish adherence to p-values. The point I’m making is, if anything, the opposite: the fact that a paper reports an interesting effect at p < .05 is *not* a sufficient reason to accept the paper, and reviewers should often feel free to ask for more data if the effect seems wildly implausible. The point holds whether you’re taking p values, effect sizes, or confidence intervals as your metric of plausibility, and I’m with you in thinking we should privilege the latter.

Comment on the truth is not optional: five bad reasons (and one mediocre one) for defending the status quo by Rich

Rich — Fri, 15 Mar 2013 00:59:58 +0000

I agree with the sentiment that the degree of evidence should be proportional to the strength of the claim, but I don’t like the implication that we become slaves to an arbitrary p-value. As long as journals only publish p < .05, we will see ad hoc explanations presented as a priori, and failure to correctly adjust for multiple comparisons.

Rather than advocating for stricter p-value enforcement, we should be asking scientists to provide confidence intervals or estimates of effect size.