The Ashley Madison Breach comes to mind. If the core demographic cares about not wanting their data on the platform to get out, they will vote with their feet. That said, I think this example is not the norm, and most people probably won't care for most applications.
https://en.wikipedia.org/wiki/Ashley_Madison_data_breach
> Have fun explaining to your spouse why your household's TV is showing more dating site ads than that of their friends.
The targeting is sometimes only _so_ good. While it works in aggregate well, sometimes there are laughable targetings, and it also depends on what audiences the advertiser wants to hit.
I get some pretty strange targeted ads through Facebook that don't seem at all relevant, and the "Why Am I Seeing This Ad" dropdown has very nebulous explanations ("Targeting Men between 25-35 in San Francisco"). I would imagine this problem to be similar with targeted ads on Live TV unless they keep the ads pretty generic or have vastly better targeting than FB or Google.
> I feel like this argument should be a class of fallacy. Lots of things have "existed" for a long time, but that doesn't mean previous iterations were effective.
This is pretty close to Survivorship Bias (https://en.wikipedia.org/wiki/Survivorship_bias). Outcomes of midwifery of yore is no different than present midwifery if you ignore the fatalities and only focus on successful deliveries.
While it is a policy issue at its core, changing law and policy moves at a glacial pace, and it's not even a certainty that it'll get changed at all (the "nothing to hide" defense gets brought up a lot on these matters, and it's a pretty persuasive argument to those that can't recognize the fallacy). Technical solutions have the benefit of being much quicker to enact, albeit in a flawed way that is, if highly successful, a band-aid on a bullet wound.
Sure, using hypothesis tests could pick out some of the structured examples in the Datasaurus, but in practice, things are often more subtle. Goodness of Fit tests to check for normality, in particular, are a little bit thorny, lacking power in small sample sizes, and rejecting normality for slight departures in higher sample sizes. My experience has been with assumption checking that by the time a hypothesis test has sufficient evidence to reject an assumption, you'd usually be able to see it visually.
Until you get into high dimensions, it probably doesn't hurt too much to visualize the data. Additionally, it can be helpful to understand what signal has been left in the residuals (ex: you fit a linear model, but failed to include a quadratic term), which is something hypothesis tests aren't as good at telling you.
> OLS works fine in classification problems. And it has advantages.
Do you have more explanation of these advantages? I read through the link you sent, and a bit more about linear probability models. Such things were never discussed in my statistics curriculum (BS, MS, PhD), except for motivating why logistic regression was necessary. I'm not sure I understand the economist's arguments in favor of LPM. Both the interpretation and the distribution of the test statistics will be totally different with OLS versus Logistic Regression, and the overall probability of a defunct project is pretty small ( \hat{P(y=0)} = .07 )--enough where there would be pretty big differences. To be clear, my reservation is with the p-values in the OLS model, not the predictions it generates. While the models agree on the direction of the covariates, the magnitudes are quite different, even when you convert logit/probit to be on the same scale as LPM.
> Multicollinearity refers to perfect multicollinearity.
Perfect multicollinearity will definitely mess up the estimation, but even if Score and Comments are not perfectly collinear, it's difficult to talk about each one's effect on the probability individually, as is the interpretation of coefficients in a (logistic) regression. What does the VIFs look like for Score and Comments, in particular?
> I mention R² as one measure of predictive power.
But the outcome is binary, so you'll have a similar issue as Minimaxir's first point about OLS. If you wanted to talk about prediction accuracy, what about a confusion matrix, misclassification rate, or specificity/sensitivity/F1? Granted, you'll not want to predict on the same tagged examples that you trained the model on, but maybe you could split it 80-20? Or tag another 20-50? There are also R²-like measures you can use when the dependent variable is binary (a whole class of pseudo-R² measures).
I would be curious to see the relationship between these predictors and the response. It's usually been my experience that linearity is a strong assumption to make, and that I'd expect for something like comments or score that once it reached a certain threshold, there was no extra value added by getting more comments/score. Are the log-score and log-comments linear over their entire support?
What about something ala Neovim? I've only ever looked at Tex from the perspective of a user (I don't program my own macros too often), so I don't know how hard it'd be, but why not a language overhaul?
I use SAS professionally at my job, and R in all my academic/hobby work. R has a couple packages that give similar functionality as PROC SQL (about 95% of my SAS workflow, since it's far nicer than data steps for a lot of things). There's an ODBC package (RODBC), as well as SQLDF, which allows you to use SQL queries to manipulate data frames in R.
While there is (almost?) always a way to do a SQL query using idiomatic R, I have to admit that sometimes my brain thinks up a solution in SQL faster (a product of upbringing).
>What kind of negative effect might result from a bunch of unqualified high school teachers teaching CS poorly? Is some exposure better than none regardless of teaching quality?
Is this problem similar enough to math that we can draw on data about mathematics education in schools? You can make a lot more money in industry with a math degree than you can teaching math to grade schoolers.
Certainly, we hear tons of stories about unqualified math and science teachers, and all the harm they do to a desire to learn math, but I'm sure some students get exposure to math that wouldn't normally have much desire to immerse themselves in it.
I agree it's a barrier to have such crazy prices, but there are free resources available, especially on a topic as popular as Grammar of Graphics. Hadley Wickham (in my mind, synonymous with the concept, since he implemented Wilkinson's ideas in R's ggplot package), for instance, has numerous materials on it, including a short primer (http://vita.had.co.nz/papers/layered-grammar.pdf). It might not be as exhaustive as the Wilkinson text, but surely there's enough material out there to implement GG in JS, especially considering there's successful implementations in other languages?
Can anyone comment if the GM fixes the battery problems Beta 1 had? I was on the first beta awhile back, and my 2012 rMBP got about an hour battery life (I usually get closer to ~3-4) and was always hot to the touch.
The same thing happened to me with Mavericks, so I chalked it up to Apple not optimizing the early betas for battery life.
Tufte-Latex, mentioned in the article, is a really nice template that produces some gorgeous Latex handouts with very little effort (I've used it a couple times when I wanted something to not look like the standard Latex article class). I thought it worth linking here for those that didn't read the link: https://tufte-latex.github.io/tufte-latex/
Fixxer, I totally agree that it's great that one be capable of doing these things, but sometimes it's not as important as other things that could be taught. Like acbart, sometimes I want to teach why/when to use a statistical algorithm and not teach them how to grab all the dependencies, troubleshoot whether they have gfortran installed, etc. This problem is horribly compounded teaching undergrads when you have Windows, Linux, and Mac users in your class, where the procedures for getting a working scientific stack vary, and the errors are often not the same across platforms.
When I install the scientific python stack on a new machine, I almost always just use Anaconda. I already know how to install the stack (I'll always value the weekend I spent in undergrad fighting with a customized BLAS in R!), but sometimes I have more pressing/fun things to do.
With Anaconda, I just tell my students to download a quick installer, slap on iPython or PyCharm, and it's ready to go. It's one less thing to worry about! The installation is dead-simple, and is almost exactly the same whether on Mac, Windows, or Linux. When I do data science, I don't want to have to be doing IT, too!
Part of the problem is that data science doesn't have nearly the same formalism in its definition that statistics does. What's the difference between BI's, Data Miners, Data Analysts, Data Scientists, etc? The tools used to arrive at conclusions (R vs. Python vs. SAS vs. Tableau/Excel/SPSS) doesn't seem like a good way of differentiating the roles.
A more useful discriminator would be the application of statistics (BI vs. Biostatistician, for instance), the depth and complexity of the statistical algorithms used, and whether the main use is stat inference or prediction (machine learning doesn't seem to focus on inference a whole lot, for example).
I come from the world of biostatistics, where diagnostic tests are usually measured in terms of Sensitivity (probability of Predicting Evil, given actually Evil, same as Precision) and Specificity (probability of predicting Not Evil, given actually Not Evil). What is the reason for choosing to balance Precision/Recall, versus Sensitivity/Specificity? They are definitely similar, so why prefer one tradeoff versus the other?
They hint at withholding data ("The same logic applies when it comes time to splitting our data into training and validation sets"), though they don't outright mention cross-validation. Seems like a bit of an oversight for an article about building ML models, I agree, but the article seems like it's mostly surface-level.