Talk:Path analysis (statistics)

Path analysis as a statistical tool requires a more detailed definition. (Person "199.196.144.13" from below fixed the typos in this one-sentence submission by another person, but person "199.196.144.13" agrees with this sentence. Meanwhile, here is person "199.196.144.13"s submission...)

Typographic error: failure to close parenthetical:

(the two dependent or 'endogenous' variables

should be

(the two dependent or 'endogenous' variables)

Substantive comments: While correct, what is written is too compact. Furthermore, it should be made explicit that Path Anlysis REQUIRES that one EXPLICITLY DEFINE ONE'S MODEL -- ALL THE ARROWS, ETC. This is done a priori. Path analysis then gives the magnitudes. Model plausability is secondary. ("The magnitudes of what?", you ask. Of direct effects, indirect effects, spurious effects, unanalyzed effects. "What are all these?", you now ask. Things that should have been discussed in the article but were not.) Plausability, though secondary, can be appraised in any model that does not have direct connections "down the line" from every variable preceeding (I believe this is called a fully recursive model). Plausability is appraised, in part, by comparing observed correlation to model-predicted correlation for all the variable pairs (unless fully recursive in which case plausability appraisal is not possible). (Plausability appraisal also relies on subject-matter expertise, not just the correlation comparisons.)

Additional substantive comment: An example would've been nice. Here's one. A sociologist believes that whether the son goes to college or not is not merely a matter of family's income. The sociologist posits that yes, income is a direct effect, but there is additionally another direct effect: if the father went to college he encourages his son to go also. So while the family may have lots of money because the father went to college, the father going to college is NOT ONLY an indirect effect in the having of the bucks (money) to send the kid, but the fact itself affects the son going, as in: "Son, I went, so you are too." This example is here in this discussion just as text; I didn't draw the diagram. But the modeling of path analysis IS both text and diagram -- though you use the diagram for computing. But what I wanted to do is give the SPIRIT of path analysis. In my example, admittedly, go-to/no-go-to college is dichotomous, so other multivariate techniques could be used, but it made for an excellent example. A furtherance of this example is intriguing: do the similar model for a daughter as well as a model for a daughter without the farther or mother going to college being a direct effect but only an indirect effect (via the family's money). Thus we could (doing both models for the son too) compare all four models. We may find that yes there is a "Son, I went, so you're going to college" effect but there is no "Daughter, I went, so you're going" effect. That is, for the daughter, it's all a matter of money, but for the son there is an additional component: "the I went, so you're going too" component. Now these considerations REALLY give the flavor of what path analysis is all about.199.196.144.13 (talk) 20:57, 27 October 2008 (UTC)199.196.144.13 (talk) 21:03, 27 October 2008 (UTC)199.196.144.13 (talk) 21:13, 27 October 2008 (UTC)

• to 199.196.144.13: I read your example with interest. Please do submit a path-diagram for it. I think the succinctness of the article itself is a virtue. It is simple, but it is correct. I think the whole piece needs expansion, including fuller mathematical development. This is, after all, a very powerful analytical tool. My interest has been through Sewall Wright in his "Genetical" guise, and his contribution to Quantitative Genetics has been monumental. Donach (talk) 03:50, 21 August 2014 (UTC)

"Path analysis (statistics" listed at Redirects for discussion

The redirect Path analysis (statistics has been listed at redirects for discussion to determine whether its use and function meets the redirect guidelines. Readers of this page are welcome to comment on this redirect at Wikipedia:Redirects for discussion/Log/2024 April 8 § Path analysis (statistics until a consensus is reached. Utopes _{(talk / cont)} 00:56, 8 April 2024 (UTC)