# A pair of ambiguous tags: [nested] and [non-nested]

We have with 216 Qs and also with 22 Qs. Neither tag has a wiki excerpt. Both are terribly ambiguous.

AFAIK, "nested" can refer (and this tag is being used to refer) to various conceptually different things that have little to do with each other:

1. Nested data / nested experimental design (nested factors in ANOVA, nested random effects in mixed models, etc.)
2. Nested models (e.g. doing LR test on two nested models)
3. Nested cross-validation

Update April 2016: I looked through the last 50 threads tagged with [nested]. 39 are about nested data (split ~50:50 between ANOVA and mixed models); 1 is about nested models; 10 are about nested cross-validation. Among the [non-nested] threads, most seem to be about nested models.

I suggest to split these two tags into three new tags, corresponding to the above categories.

Questions:

1. Would [nested-data] be an okay name for nested anova/lmer stuff? We could have nested-anova, nested-random-effect, nested-factor as synonyms. There is also a [multilevel-analysis] tag, but people asking about nested anova do not usually use this terminology...

2. [nested-models] would do, wouldn't it?

3. Do we need [nested-cross-validation] tag, or can we retag these threads to [cross-validation] and that's it?

• I upvoted the question. I'm not sure what I think the answer should be. The problem is that "nested" isn't an noun phrase, it's an adjective. So we either end up with as many different [nested-____] tags as can exist, or we have a single [nested] tag that means something different depending on the other tags applied. – gung - Reinstate Monica Oct 5 '16 at 23:45
• @gung That's an interesting perspective. Do you know any other adjective-like tags that can be combined with various other tags to refer to different concepts? I am wondering if this is something that can work well, or should rather be avoided. – amoeba Oct 5 '16 at 23:53
• I don't know. It really needs some hard thought, but I've never really had a chance to work though it. – gung - Reinstate Monica Oct 5 '16 at 23:55
• Bumped again ;-). – gung - Reinstate Monica Nov 4 '16 at 14:09
• @gung After it was bumped again, I made an update with some usage statistics. – amoeba Apr 3 '17 at 21:28

My proposal:

1. Re-tag usage #1 to :

Use this tag for nested random effects in mixed models and nested factors (or nested design) in ANOVA.

2. Re-tag usage #2 to :

One model is "nested" in another if it is a constrained version of it. Nested models can be compared with a likelihood-ratio test. Use this tag for questions about comparing non-nested models too.

3. Remove usage #3. I don't like [nested-cross-validation] tag too much because it would be a proper subset of ; also, most existing threads about nested CV are not using [nested] tag anyway.

Update (May 5): [non-nested] is gone.

Update (May 8): I have moved all threads that are not about [nested-data] out of [nested]. The remaining 172 questions can now be moved into [nested-data] via mod action.

Update (May 9): This is done, see comments. Case closed.

• presently merging [nested] into [nested-data] – Glen_b May 8 '17 at 10:21
• I believe it's complete. – Glen_b May 8 '17 at 10:22
• @Glen_b Thanks! If possible, could you now remove the synonym mapping nested-->nested-data? There is no sense in automatic mapping of nested into either nested-data or nested-models in the future. – amoeba May 8 '17 at 19:13
• I have some misgivings about that (such as people now using "nested" again), but done as requested. – Glen_b May 8 '17 at 22:34
• @Glen_b Thanks. I would not worry about reappearing of "nested". So far, whenever we replaced tag A by A-foo and A-bar (I think this happened a couple of times), A did not reappear. I think this makes sense because when people start typing A, they get A-foo and A-bar as suggestions and will choose one of them. – amoeba May 9 '17 at 6:50

You apparently are identifying a very nested topic. I have read about either well-or ill-defined other uses: , (nested analysis of variance or covariance with nested nominal variables), (of which hierarchically structured data are a subclass), .

So it definitely makes sense. Could this process lead to the need of a tag ontology?

References:

• Hmm, nested-data might actually be a better name for my point #1 (nested-design or nested-effects). Nested-anova is anova with nested factors (isn't it?), so it's the same thing. What is nested-analysis? If you mean analysis of nested data, then it's the same thing too. Can't say anything about "nested sampling" because I don't know what that is... – amoeba Sep 2 '16 at 23:21
• I think [nested-data] can cover nested ANOVA, nested ANCOVA, nested factors, nested experimental design, nested random effects in mixed models, and also "nested analysis" (whatever it is). Don't you think? – amoeba Sep 5 '16 at 12:49
• @amoeba I think nested-data is a bit too data-centric: if you do not like nested-models or nested-analysis, perhaps nested-processing? I do not fear the confusion with parallel processing as we are on SE.stat – Laurent Duval Sep 5 '16 at 12:56
• Wait, I do like nested-models, see point #2 in my answer. But when one has nested factors in anova, it is because the dataset is nested itself; it's not a choice that an analyst does, the analysis has to incorporate nested factors if they are nested; that's why I like nested-data. – amoeba Sep 5 '16 at 12:59
• @amoeba The choice of the grouping, the depth of nesting can be user dependent, isn't? I tend to believe that the act of choosing data, the way they will be processed, is a part of the processing (as it involves interpretation), but one could call that an opinion – Laurent Duval Sep 5 '16 at 13:46