Talk:Observational study

STROBE
STROBE is an international effort to improve the validity of observational studies:
 * Explanation and elaboration
 * Guidelines for reporting

This really ought to be mentioned. JFW | T@lk  17:21, 16 October 2007 (UTC)

Merge natural experiment into observational study
The subject of both articles appear to be where the assignment of subjects into a treated group versus a control group is outside the control of the investigator/"experimenter". Neither article is long enough to justify being separate for that reason. Observational study is, IMPOV, the most appropriate target, since the term experiment usually implies controlled experiment, which is, by definition, distinct, so the term natural experiment is somewhat contradictory.Mikael Häggström (talk) 06:44, 7 September 2010 (UTC)
 * I would not be opposed to the merger, but would like to point out that "natural experiment" is used often in the geosciences to denote a naturally occurring process of interest for which many (or most) variables are well-constrained, and can provide insight into the broader physics or chemistry controlling the process (thus making it generalizable). In spite of what you say about "experiment", "seismic experiment" is used for a deployment of seismometers to image a portion of the Earth (this bothers me too). Looks like I should actually get to editing this article at some point... in any case, a merge would be just fine. Awickert (talk) 06:51, 7 September 2010 (UTC)
 * I suggest that you need to look at other articles that link to these two articles and where trouble has been taken to make a distinction between the meaning of the two terms. Melcombe (talk) 08:33, 7 September 2010 (UTC)


 * (copied from Talk page of other article) I am opposed to such a merger. The natural experiment is a rare and unusual type of study. It is observational, yes. But it is the rare kind of observational study that provides a basis for ascertaining causality. I think it was wise to raise the issue because the issue needed airing. However, I think that the natural study should stand by itself, with links back-and-forth with the observational study entry. I plan to include such links. Iss246 (talk) 14:18, 7 September 2010 (UTC)
 * I would support the merger. I am a psychologist and one cannot allocate causality from a natural experiment or observational experiment. The definition given on this website of a natural experiment is what an observational experiment is known as in psychology. 10:09, 3 November 2010‎ 62.172.145.195(talk)‎

The term natural experiment originated in economics, although it has spread. The term natural experiment has a precise meaning that is distinct from what is normally meant by an observational study. In a natural experiment, the treatment (independent variable of interest) is driven by something exogenous to the dependent variable; thus, it mimics what a researcher would want to do if they could do an experiment. It contrasts with almost all ordinary observational studies, where the treatment (independent variable of interest) is endogenous, making it extremely hard to determine a causal effect.

For discussion, see (among many other references) "Mostly Harmless Econometrics" (http://www.mostlyharmlesseconometrics.com/) or chapter 13 on Natural and Quasi Experiments in the book "Research Methods in Practice: Strategies for Description and Causation" (http://www.sagepub.com/books/Book232420). (Full disclosure: I co-authored the latter book.)

The natural experiment is an important sub-topic and should not be moved to observational study.Dkremler (talk) 19:29, 5 March 2011 (UTC)dkremler


 * I totally agree with Dkremler. The concept of natural experiment emphasises not only that opportunities are present "in the wild", but a particular type of opportunity and scientific practice where one finds conditions that match those of a well-controlled experiment, e.g. villages that are matched for size and economic status, but one or a few are about to have wells introduced by a charity.  The researcher must find out about the opportunity (in this case, 3rd party intervention) then find the well-matched comparison (the naturally occurring controls, in this case the most similar other unaffected villages.)  This example is due to work done by the group of [Ruth Mace] --Joanna Bryson (talk) 18:44, 7 May 2012 (UTC)


 * I'll remove the merge templates as they are old and there is certainly no consensus that it should be done. But can you (Joanna Bryson or someone) include something along the line of the two contribs above to clarify the difference. Melcombe (talk) 21:29, 8 May 2012 (UTC)

Inconsistency about fields
The introduction says this article is about the term as used in epidemiology and statistics. The remaining part of the article mostly seems to describe therapy trials while the second to last paragraph mention its use in the social sciences and medicine in general. Which is it: epidemiology and statistics or a broader range of fields? Perhaps the first part ("In epidemiology and statistics,") should be removed.

I think that if it's used in a broader range of fields the article should cover the entire term (moreover, the article mostly seems to be about therapy trials, so it's not about epidemiology and it's not about pure statistics but it is biased towards a single subject). The sources for the first sentence seem to be an epidemiological source, and a general medical source, the latter is about the term as used in research about diseases and treatments in general, indicating it is indeed used in those fields.

As a last note I'm wondering if there are any more fields where it is used, for example: Ruxton and Colegrave mention it in Experimental design for the life sciences (2nd ed.) without reserve to which subfield it is used in. Moreover, the example they name is about natural selection. They do call it correlational study, but they mention it is synonymous with observational study. However, there may be more fields in which it's used, although I'm not sure if a list of fields needs to be added to the article at all. PinkShinyRose (talk) 09:10, 6 October 2013 (UTC)


 * Perhaps we should consider replacing the whole intro paragraph with something more generally applicable. Observation, and observational studies are used in all fields of inquery.
 * Perhaps something like this, I understand that it would need editing:
 * "An observational study is a data collection method used in all branches of scientific inquery. In observational studies, researchers observe the phenomena of interest without changing them. Observational studies may be contrasted with experiments, where researchsearchers impose changes and observe the results. Observational studies may demonstrate correlation but not causation."
 * Ammon0 (talk) 02:27, 16 October 2014 (UTC)
 * I edited the first section to more accurately reflect the scope of observational studies -- PeterLFlomPhD (talk) 21:57, 17 August 2015 (UTC)

Scope far too limited
This article makes it sound like observational studies are mainly relevant to social sciences and epidemiology -- fields at which hard scientists often smile. This is to omit massive swaths of science in fields like: astronomy, astrophysics, archaeology, anthropology, paleontology, zoology, and possibly much of biology. I don't understand why these are ignored.

Quality section citation
For the study cited under the quality section(2014 Cochrane review), I’m not sure if the findings of that study is properly communicated:

1. There’s no mention that it is specifically about healthcare outcomes 2. (I might be wrong here) my understanding of the study was that they didn’t compare inferences made from observational studies to the outcomes of controlled studies. Their main finding was that the observations made in observational studies were usually inline with the expectation predicted by randomized controlled studies. I think there’s a huge difference between doing a controlled study, then an observational study to confirm expectations with little error, and doing an observational study and making inferences with no error. To illustrate my point, if I do a proper controlled study of a diabetes drug, and then release it to public, there’s a very high chance that observational studies will confirm the effects of that drug. On the other hand, if I was conducting an observational study of a population, trying to figure out if consumption of a drug can treat diabetes, there’s a good chance there’ll be many false positive, due to bias, behavioral correlations, etc.

I’m not sure how to rewrite it to reflect those facts, but I feel like the paper made those clear while the wiki was very misleading 74.101.22.170 (talk) 19:41, 25 January 2022 (UTC)

Ja
In fields such as epidemiology, social sciences, psychology and statistics, an observational study draws inferences from a sample to a population where the independent variable is not under the control of the researcher because of ethical concerns or logistical constraints. One common observational study is about the possible effect of a treatment on subjects, where the assignment of subjects into a treated group versus a control group is outside the control of the investigator.[1][2] This is in contrast with experiments, such as randomized controlled trials, where each subject is randomly assigned to a treated group or a control group. Observational studies, for lacking an assignment mechanism, naturally present difficulties for inferential analysis. 49.145.99.16 (talk) 13:52, 14 January 2023 (UTC)