Wikhypothesis

This is a proposal for a new Wikimedia sister project.

Wikhypothesis
Status of the proposal
Status	rejected
Technical requirements

Wikhypothesis

Whereas Wikipedia is a repository of facts, Wikhypothesis is a repository of hypotheses about how to change outcomes and all relevant evidence for said hypotheses. Thus, Wikhypothesis could serve as the backbone for a metaanalysis of any hypothesis humans care to consider by becoming an extremely concise metajournal of all human knowledge and knowledge acquisition.

Some example hypotheses:

Does a change in X	have a significant impact upon Y?
Newborn sleep position (front vs back)	Newborn's morbidity and mortality rate?
Tax advantages of a Roth IRA in the United States	Gross household savings in the United States?
Quantity of lye-based drain clog remover used per year	Lifetime depreciation of the relevant plumbing?
Quantity of biological children	Female politician's probability of winning an election in the Nordic nations?
Population size of the American black bear (Ursus americanus) [1]	Dollar value of downstream flood damage?

See Sample Wikhypothesis page below

Goals of Wikhypothesis

The five areas of focus illuminated for any possible hypothesis would be:

• Methods
• Models
• Mechanisms
• Conclusions
• User Experience

Methods

Which methods were used to acquire data and/or conduct study?

• Neutrality: This project is neutral as to methods and simply reports any method and any limitation without value judgement. For each hypothesis page, only coded elements in the table receive a reference. While there may be 1,000 methods to knowledge, each page does not need to list them all.

Models

Which mathematical, analytic, conceptual models were used to organize and analyze the data?

Mechanisms

What is the proposed mechanism of action, how does X -> Y?

Conclusions

What is the basic outcome of the study: to accept or reject the hypothesis?

User Experience

How to make Wikhypothesis work for users?

• Summary: Wikhypothesis must offer a metanalysis of all the evidence for and against each hypothesis in order to quickly help the user decide on their own confidence that the hypothesis is, or is not, true.
• Filtering: Wikhypothesis should introduce a filtering function for each column. Suppose you wanted to see only the studies using methods "a" or "m," you should be able to filter out the rest.

Sample Wikhypothesis page

• Daily Intake of Selective Serotonin Reuptake Inhibitor XYZ --> Symptoms of Clinical Depression?

Summary: The Hypothesis is Very Likely True

Study	Year	Source	Funding	Methods	Models of Analysis	Conclusion	Limitations	Proposed Mechanisms	??
Study 1	2013	Nature	NIH	a, m	c, e	Yes	c, f2	Chemical: neurons	Example
Study 2	2011	Lancet	OECD	META	a, c, e, f, m, n, o	No	x, j19	Chemical: neurons; Publishing bias	Example
Study 3	2010	JONEM	NIH	a, e, m	c, e, f	Yes	c, f2, p3	Chemical: neurons	Example
Study 4	2009	Unpublished	??	a	h	None	c	Chemical: neurons	Example

Methods:

• a: large-N sample size link
• e: large-N sample size double blind link
• ..etc.

Models of Analysis:

• a: linear regression least ordinary squares link
• f: logit model link

Proposed Mechanisms:

• c44: Chemical: neurons link
• mi23: Publishing bias link

Proposed by

@jtdjtdjtdjtd