Reputation
10,413
Top tag
Next privilege 15,000 Rep.
Protect questions
Badges
1 14 51
Newest
 Nice Answer
Impact
~267k people reached

2d
revised How does the R function arima() calculate its residuals?
edited title
Jul
30
awarded  Nice Answer
Jul
30
comment Why would parametric statistics ever be preferred over nonparametric?
... and @StasK expressed it all much better than I did.
Jul
30
comment Why would parametric statistics ever be preferred over nonparametric?
The null hypotheses are different between a parametric test and its nonparametric counterpart. Specifically, the null hypothesis for a parametric test contains a specific parametric assumption on the distribution of the test statistic (which will usually also be calculated in different ways for the two tests) - that's why it's called "parametric", after all! So the two p values have the same name, but are calculated based on different test statistics, which have different distributions under different null hypotheses.
Jul
30
answered Why would parametric statistics ever be preferred over nonparametric?
Jul
29
answered Forecasting methods for monthly sales
Jul
28
comment Should parsimony really still be the gold standard?
+1. I suggest reading The Elements of Statistical Learning (freely available on the web), which discusses this problem in depth.
Jul
28
comment how to modify box plot diagram
This does not look like it was produced by ggplot2. In base graphics, look at the pch parameter. Look at ?points to find a list of possible point shapes. (These are not boxplots.)
Jul
28
answered Nonlinear forecasting
Jul
27
comment Why don't log-likelihoods lead to log(0)?
@marcman: of course pixels can have the value zero. But the probability model you use (and from which you calculate the likelihood) should not assign a zero probability for a pixel to have the value zero.
Jul
25
answered Benchmarking time series forecasting model
Jul
22
reviewed Leave Open Distribution of sum of function of two random variables
Jul
22
comment Distribution of sum of function of two random variables
Any information on $p$ or $f$? (I don't assume $f$ is symmetric, is it?)
Jul
22
comment Why doesn't deep learning work well with small amount of data?
+1. Model complexity should always only grow slowly with sample size, and deep learning is a pretty complex model, implying that it will usually not work well for small sample sizes. The Elements of Statistical Learning (available for download for free) discusses this - highly recommended.
Jul
22
accepted Sum of squared Negative Binomial probability masses
Jul
22
accepted Sum of squared Poisson probability masses
Jul
21
revised Sum of squared Poisson probability masses
added 168 characters in body; edited tags
Jul
21
revised Sum of squared Negative Binomial probability masses
added 46 characters in body
Jul
21
comment Sum of squared Negative Binomial probability masses
Thank you. This looks both intriguing and intimidating. I'll need a little time to digest it. In the meantime, could you please introduce me to your friend ${}_kF_\ell$?
Jul
21
comment Sum of squared Poisson probability masses
Thanks, this is very helpful. What would be the advantage of using gsl::bessel_I0() rather than base::besselI(nu=0)?