# VERY ambiguous [population-average] tag

Have you ever wondered what is about..? If not, check its info page

Population average model assesses the association between X and Y based on averaging over the random effects, rather than the one for an individual subject (subject specific model).

Somehow it does not surprise me that it is not used for such questions...

• That sounds like a reference to the GEE. – gung - Reinstate Monica Oct 10 '16 at 12:29
• @gung maybe, but it's a guessing game... – Tim Oct 10 '16 at 12:32

That sounds to me like a reference to the GEE, although it is ambiguous. (Some of the threads it's used on support this interpretation.)

I think this tag's name will inevitably cause it to be used incorrectly by the many statistical novices who ask questions here. I don't immediately see how that can be addressed effectively. For example, it has an excerpt, but many people don't read them anyway. Thus, I don't think making it a synonym of something else is the right option. (We'll just end up finding some other tag commonly misapplied.)

I think the best we can do is probably just to remove the tag from threads and let the SE system purge it after 24 hours. An alternative would be to replace it with a new tag that has a different name that won't be misunderstood by novices. I'm not sure what the appropriate name would be, though, and existing tags, such as , might be enough already. The tag may have been created by @JiebiaoWang; it would be worth seeing what he has to say. I attempted to leave him a comment alerting him to this thread.

Update: It seems [marginal-model] might be a good name for a new tag to put in place of the old one (h/t @JakeWestfall). I propose that we work through the threads currently tagged with and, where appropriate, replace it with the new tag. Where the tag had been misapplied, it can just be removed. Once it doesn't exist on any threads, the SE system will automatically purge it in 24 hours.

Update 2: Given the response to this, I will go through the threads tagged with [population-average] and remove the tag. I may replace it with [marginal-model] and/or [gee] (or something else) as seems most appropriate. I have now done the first such thread (here).

Update 3: I have now removed or replaced the tag on all threads. The tag should disappear after a day or so.

• +1; It seems that this tag is confused with "population MEAN", though the info clearly states that it means population average models. Searching "population average" (with quotes) in Google shows the correct results. – Randel Oct 11 '16 at 5:24
• @JiebiaoWang, can you think of a name that might be more distinct / less confusable for novices? – gung - Reinstate Monica Oct 11 '16 at 14:17
• How about "population average model"? I also agree to remove this tag if it is easier to do so. – Randel Oct 11 '16 at 15:22
• How about "marginal model"? – Jake Westfall Oct 11 '16 at 17:14
• @JiebiaoWang, would you be OK with [marginal-model]? That doesn't seem likely to be picked by novices accidentally. – gung - Reinstate Monica Oct 11 '16 at 17:58
• Seems perfect for me! – Randel Oct 11 '16 at 19:05
• marginal-model is not really a new tag; it exists already. I don't know anything about this topic, so I have no idea how it is related to [gee], and tag excerpts do not help me here. There is also [marginal-estimate] without an excerpt and also simply [marginal]; the latter is probably way too ambiguous and might require some attention. – amoeba Oct 11 '16 at 20:24
• @amoeba, hmmmm. I'm don't have as much expertise here, but I wouldn't think of marginal models being synonymous w/ the GEE; I would say the GEE is the most prominent case of MMs (eg, see the end here). The excerpt / wiki could probably some attention. W/ that done, MM might still be the best option. – gung - Reinstate Monica Oct 11 '16 at 20:45
• @amoeba, there are MMs that are not GEE, but they are uncommon. Using [gee] alone is probably appropriate most of the time. We could even put in the excerpt 'for MMs that are not GEEs', or some such. We again have the issue here that is analogous to all [t-test] is a subset of [regression] is a subset of [gen-lin-mod], etc. – gung - Reinstate Monica Oct 11 '16 at 20:52

I would suggest to rethink its existence or use with respect to the tag (which seems undescribed) and a novel tag, like or .