From 19315eb015b6a4493a4efad36129a7d284a9996c Mon Sep 17 00:00:00 2001 From: David Strayhorn Date: Mon, 29 Jan 2024 18:28:30 -0500 Subject: [PATCH] Update calculationsOverview.md --- .../curatedLists/v021Overview/calculationsOverview.md | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/appDescriptions/curatedLists/v021Overview/calculationsOverview.md b/appDescriptions/curatedLists/v021Overview/calculationsOverview.md index d9fc440..976c2f4 100644 --- a/appDescriptions/curatedLists/v021Overview/calculationsOverview.md +++ b/appDescriptions/curatedLists/v021Overview/calculationsOverview.md @@ -18,7 +18,9 @@ Each rating is also associated with a "confidence" score, which is a number betw The goal is to calculate, for each user, a quantity called *influence* which is used to determine that user's weight. Each list is its own independent "context" -For each user, an *average score* is first calculated, which will be a number between 0 and 100. The average score is a weighted average, calculated from every trust rating. The weight of each rating is equal to the *influence* of the rater, multiplied by the confidence score of the rating (set to 0.8 for all ratings for now), and multiplied by the *attenuation factor*, which is a user-controlled parameter between 0 and 1. The default trust score is treated as its own rating that is applied to ALL users. The default trust score has a confidence that can be adjusted from 0% to 100%, independently of the default 80% confidence rating for regular ratings. +For each user, an *average score* is first calculated, which will be a number between 0 and 100. The average score is a weighted average, calculated from every trust rating. The weight of each rating is equal to the *influence* of the rater, multiplied by the confidence score of the rating (set to 0.8 for all ratings for now), and multiplied by the *attenuation factor*, which is a user-controlled parameter between 0 and 1. + +The default user trust score is treated as its own rating that is applied to ALL users. The default trust score has a confidence that can be adjusted from 0% to 100%, independently of the default 80% confidence rating for regular ratings. The purpose of the dfault trust score is to define the influence score of the "unvetted user": the user who provides a rating but who has not been rated by any other user. For each average score, a variable called *input* is defined as the sum of all of these individual weights. In theory, the input ranges from 0 to infinity. This number is mapped to a score called *certainty* which is a number between 0% and 100%. The equation to convert *input* into *certainty* is an exponential, the decay of which is controlled by a parameter called *rigor* controlled by the user in the control panel. @@ -36,6 +38,11 @@ Summary: Similarly to the above. +Items are sorted into bins: +- PENDING: items whos input is below some threshold (not enough ratings) +- ACCEPTED: +- REJECTED: + # Screenshots The PGA desktop app lets you review the raw data and all calculations.