Talk:Longitudinal study

From Wikipedia, the free encyclopedia

[edit] Correlation

I notice that someone has changed the last paragraph to say that longatudinal studies are "observational". I would like someone to explain what the difference between an observational and a correational study is.

A correlational study is a study that looks at relationships between two variables without "manipulating" any of the variables. Which is excactly what a longatudinal study does.

Ivan 05:19, 23 February 2006 (UTC)

I don't think you're incorrect, but I think that "observational" is a more widely used (and better) term. To me, "correlational" hints too strongly to a particular measure of association, whereas I think the experimental/observational pair picks up better the contrast between experimental manipulation of the world, and passive observation of the world. It's also widely used terminology; to hand I have Rosenbaum's Observational Studies, Springer 2002, which is a very interesting text on statistical techniques that use experimental reasoning on observational data.

I will expand a little on the term in the article.

You'll also note that I modified the claim about causality. While no data collection technique can give us causal relations (not even experiments; causality is imputed, not observed), there is a hierarchy in the extent to which methods give us information about causality: experiment > longitudinal > cross-sectional. BrendanH 15:20, 23 February 2006 (UTC)

Another data point: Causality#Probabilistic_causation uses observational in this sense too. BrendanH 15:29, 23 February 2006 (UTC)

Correct me if I'm wrong, but longitudinal research is not necessarily observational, but can take any form that's required. Longitudinal studies might be observational in some/many cases (common example: where ethical considerations prevent us from carrying out longitudinal experiments, such as whether smoking causes cancer. Thus, we just oberve those who freely choose to smoke.), but there's no requirement that longitudinal studies are "observational", per se. For example, they might involve the use of a particular rehabilitation scheme: control and experimental groups can easily exist under these circumstances. Admittedly, both groups may be observed after being subject to the treatment/experimental conditions, but this form of "observation" is entierly different to an "observational design". I'm a little concerned the current article mightn't highlight the potentially experimental flavour, and the naive reader may incorrectly understand longitudinal studies as being restricted to an observational nature. Any thoughts? Skittled 02:05, 26 May 2006 (UTC)

Are you positive that longitudinal studies are defined to be correlational/ observational? I remember pretty clearly to have learned once that they *usually* are (and yes, that some believe they oughta be...), but no, that they do NOT have to be. Which (to me) makes perfect sense because otherwise a term would be missing for "longitudinal experiments" (which, according to your definition, would be a contradiction in terms). Or how would you call those? Thanks. (Peter)

[edit] rewording

Can someone reword "...by virtue of being able to exclude time-invariant unobserved individual differences" ? I'm having trouble understanding that... (or at least, expand on it... it looks like it's a concise definition, but with all the negatives and invariances in there... it's hard to tell what it means.) Thanks —Preceding unsigned comment added by 130.254.194.238 (talk) 17:22, 30 October 2007 (UTC)