Wikipedia:Articles for deletion/Circular analysis

Circular analysis

(Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL)

Contested PROD: "Appears to be a neologism used in this sense only by Kriegeskorte and his collaborators to describe a longstanding habit in scientific malpractice." Illia Connell (talk) 17:26, 31 March 2013 (UTC)[reply]

Weak keep. The term seems to be in use in sufficiently wide scholarly circles. Kriegskorte's use of the term is discussed here, for example, and a number of journal articles. Deltahedron (talk) 18:04, 31 March 2013 (UTC)[reply]

Keep The concept is a form of circular reasoning or begging the question, which has been common in this field, as discussed in Foundational Issues of Human Brain Mapping. Kriegeskorte's presentation of the issue seems influential, e.g. "Kriegeskorte and their colleagues deserve unending commendation...". Warden (talk) 21:17, 31 March 2013 (UTC)[reply]

Keep. It is unclear whether this topic is similar enough to any other articles to be merged into. This [review] highlights the topic's importance and suggests it may be a version of selection bias, but it is unclear whether that article uses the term in the same way. 81.98.35.149 (talk) 21:24, 31 March 2013 (UTC)[reply]

Note: This debate has been included in the list of Science-related deletion discussions. • Gene93k (talk) 01:35, 1 April 2013 (UTC)[reply]

~~Delete~~ Merge (seems quite plausible) as original prodder; I've taken a look at such references included above as I can, and I find nothing to indicate the phrase is in wide useage. Certainly the concept is significant (and well understood to anybody who's ever observed a freshman lab at work) - I would support a merge ~~to something like Selection bias, but this does seem to have a different twist than the standard bias, and so may not be a good merge target~~ to something like Testing hypotheses suggested by the data, below Ray^Talk 02:13, 1 April 2013 (UTC)[reply]

The combination of "the concept is significant" and "does seem to have a different twist" sounds like an argument for a separate article? Deltahedron (talk) 08:16, 1 April 2013 (UTC)[reply]

I'm more concerned that this is a very old concept, certainly older than Kriegskorte's papers, and I've only seen it called "circular analysis" in recent years. Back in school, we called it "adjusting the data to fit the model." Ray^Talk 18:36, 1 April 2013 (UTC)[reply]

Keep Google scholar gives 563 citations for the nature paper[1]. Highly infulential. Also happy with a merge.--Salix (talk): 07:41, 1 April 2013 (UTC)[reply]
~~Merge~~ with Testing hypotheses suggested by the data, leaving a disambiguation page that also points to Directional statistics. Devoting an article to a new term for an old concept would give it undue weight. Statisticians have been thinking about these issues for ages - see Exploratory data analysis, Data dredging, Post-hoc analysis and Multiple comparisons. Judging by searches on Google Scholar, both "circular inference" and "double dipping" were first used for this concept in 2007 (Baker, C. I., et al. "Circular inference in neuroscience: The dangers of double dipping." annual meeting of the Society for Neuroscience, San Diego, CA. 2007). Before that, "circular inference" mainly applied to analysis of directional data - hence the need for a disambiguation page. RockMagnetist (talk) 16:39, 3 April 2013 (UTC)[reply]
Merge to Testing hypotheses suggested by the data. Keep Circular statistics is what I thought of when I saw the title. I agree that this concept has been around a long time, with some neuroscientists rediscovering it for themselves just recently. I think it is worth a mention as an alternative term in Testing hypotheses suggested by the data, but not more. Update: In the interests of consensus, changing to keep with a discussion of where to merge to take place on the talk page. --Mark viking (talk) 18:45, 3 April 2013 (UTC)[reply]

The phrase testing hypotheses suggested by the data is not a catchy title and that article is not well supported by sources which verify the usage as anything more than a general phrase. Post-hoc analysis seems more succinct but, again, the references in that article do not support the usage as a title. Warden (talk) 18:58, 5 April 2013 (UTC)[reply]

I agree that it's not a catchy title, and the article needs work. Maybe the statistical community doesn't have a technical term for it - in which case a non-catchy title that clearly describes the subject is appropriate. To keep this article is to imply that some neuroscientists have done anything more than explain a well-known concept to a new community. RockMagnetist (talk) 19:58, 5 April 2013 (UTC)[reply]
Perhaps Testing hypotheses suggested by the data should be merged with Statistical hypothesis testing (see also Exploratory and confirmatory approaches). But I don't think it's necessary to solve that problem in this AfD. RockMagnetist (talk) 20:18, 5 April 2013 (UTC)[reply]
I agree that this is not an elegant title. I'm open to other semantically similar targets, such as Post-hoc analysis. That the neuroscientists have created their own term for it is fine, but it is not mainstream outside of neuroscience. As far as I can tell from Everything you never wanted to know about circular analysis, but were afraid to ask, their conception of circular analysis boils down to two things (both mentioned in the paper, so no OR on my part): (1) ignoring corrections for multiple comparisons and (2) inflated effect estimates due to selection bias of experiments that produce at least a certain size effect. Ignoring corrections for multiple comparisons involves ignoring negative results, so is also a kind of selection bias. Perhaps selection bias would be a better target? --Mark viking (talk) 20:49, 5 April 2013 (UTC)[reply]

Comment -- I've seen the concept, but under different names and guises -- circular reasoning, intellectual dishonesty, academic dishonesty, and, in its honest use, iteration. But I don;t knwo what to do with this one. Bearian (talk) 17:39, 8 April 2013 (UTC)[reply]

Relisted to generate a more thorough discussion so a clearer consensus may be reached.

Please add new comments below this notice. Thanks, J04n(talk page) 10:36, 9 April 2013 (UTC)[reply]

Keep. Salix alba points out that Kriegeskorte's paper has 563 scholar citations (actually 566 when I looked). This is a very high number, for comparison Einstein's The Meaning of Relativity, which I would expect to be off the scale in terms of citations, gets 2232 only x4 as many. Not only that, many of the citing papers are themselves respectably cited and some of them are addressing the same subject. For instance Fiedler cites Kriegeskorte and calls this effect "voodoo correlations", which term itself gets a number of book and scholar hits. Even if there is some subject overlap with other articles, it is quite clear from the sources that Kriegeskorte's take on this is highly notable. I am not opposed to a merge, but it needs to be done in a considered manner as part of the normal editing process and not forced by an inappropriate AFD decision. Spinning Spark 12:40, 9 April 2013 (UTC)[reply]

Don't let the process distract you. By all means register your protest that this is being discussed at an AfD, but keep in mind that it has become a merger discussion. How would you respond to a normal merger proposal? RockMagnetist (talk) 14:37, 9 April 2013 (UTC)[reply]

Merge: a short subsection in Misuse of statistics might do best. The above discussion has mentioned both "Testing hypotheses suggested by the data" and ""adjusting the data to fit the model" as possible meanings, but these are clearly very different faults. It may be both ae covered by the original paper. Melcombe (talk) 16:58, 12 April 2013 (UTC)[reply]
Keep - I think a merge is the correct decision, but quite a few destinations have been proposed, and we don't seem to be converging on one. It would be better to discuss the merge separately. RockMagnetist (talk) 00:49, 13 April 2013 (UTC)[reply]