Take the 2-minute tour ×
Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

I know that this question is hard to answer, but I was wondering how you would determine if a statistical technique was in current use, or if it had a heyday a couple of decades ago but it's either been superseded by newer techniques or perhaps it was even proven to be inadequate and abandoned?

For example, I was looking at a time series with R and stumbled onto the EMD package (Empirical Mode Decomposition). The technique looks like interesting, but as I google around, it seems that it never really became popular. Was it confined to the neural research community? Did wavelets prove to be superior? Did it prove to be a bit hard to interpret?

Perhaps there can be no definitive answer, but perhaps there are signs you look for?

share|improve this question
1  
I googled the original paper, it has $5789$ citations, $1900+$ in 2010-2012. It seems to be quite popular. –  user10525 Jul 12 '12 at 14:29
    
@Procrastinator: Whoa, my google-fu is pretty bad! (Or my comprehension.) –  Wayne Jul 12 '12 at 14:37
add comment

3 Answers 3

up vote 7 down vote accepted

There's no shortcut for experience, but one way to get a hint is to look up the method's original references on google scholar and see the number of citations.

share|improve this answer
    
Good tip! I just looked up the topic, apparently found the original article, and it's referenced 11 times 2010-2012, in a couple of different fields. That's reasonable, though it might also indicate people throwing an interesting technique at a problem in their field to get a publication. As you say, experience does matter. –  Wayne Jul 12 '12 at 13:18
add comment

Many techniques are not so popular, however they may be useful for specific research purposes. If a statistical methodology is supported by Monte Carlo simulations, and if it was published on prestigious journals, so I think you can take it in consideration, even though it was little used by researchers. For example, I recently used the non-parametric ANCOVA developed by Conover and Iman, but to my knowledge it is not so popular in my research field. In my opinion, it is more important to have solid foundations about why you used a specific technique (robustness related to your data, Monte Carlo simulations...) in order to adequately answer possible referees concerns.

share|improve this answer
add comment

I agree with use4733. My planned answer was to be "ask the experts". Regarding this particular technique I personally can't say much because I hadn't heard of it before even though I have some expertise in time series analysis. However, since the references seem to all be over the most recent three years it could just be that it is a very new method. New methods in almost any field take time to catch on. So I don't think that you should assume that it is not useful. It may eventually prove to be useful.

share|improve this answer
1  
For completeness, I believe the original reference was: "The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis" by Norden E. Huang, Zheng Shen, Steven R. Long, Manli C. Wu, Hsing H. Shih, Quanan Zheng, Nai-Chyuan Yen, Chi Chao Tung and Henry H. Liu, in Proc. R. Soc. Lond. A 1998 454, 903-995 doi: 10.1098/rspa.1998.0193. –  Wayne Jul 12 '12 at 14:31
    
Even though the paper came out in 1998, 14 years is not a long time and based on Procrastinator's comment it may already be popular. Consider the bootstrap. Efron brought it out in 1977 with the first paper in the Annals of Statistics in 1979. But I don't think it got much practical use until the mid 1980s and really didn't catch on until the 1990s. –  Michael Chernick Jul 12 '12 at 14:43
2  
... and then of course there's Metropolis-Hastings, with the original Metropolis paper back in 1953 and the Hastings 'update' in 1970... –  jbowman Jul 12 '12 at 15:21
    
To jbowman's point MCMC was being used by physicists and other scientists in the 50s and 60s. But Geman and Geman introduced it to the image processing community in the 1990s and when the Bayesians recognized its use in constructing posterior distributions it took off in the statistical community like wildfire. –  Michael Chernick Jul 12 '12 at 15:37
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.