1,683
edits
Changes
Jump to navigation
Jump to search
mLine 16:
Line 16: − + + + + + + + + +
→Self Assessment
User can do the assessment right after an incident or as an aggregate of a recently passed time period. Most things user would want to measure do not occur as easily defined and measured incidents so aggregation will need to be used. Unfortunately, having to remember a state over many hours requires effort and creates ambiguity. In addition, the method of aggregation (mean, mode, extremes) loses information, take a bit of effort, and can be a cause of drift. By information loss I mean that if a weighed mean is used to represent a time series the distribution is lost. For example, pain that spikes to high values for a few minutes every hour can have the same number as just pain that stays the same throughout the time period.
User can do the assessment right after an incident or as an aggregate of a recently passed time period. Most things user would want to measure do not occur as easily defined and measured incidents so aggregation will need to be used. Unfortunately, having to remember a state over many hours requires effort and creates ambiguity. In addition, the method of aggregation (mean, mode, extremes) loses information, take a bit of effort, and can be a cause of drift. By information loss I mean that if a weighed mean is used to represent a time series the distribution is lost. For example, pain that spikes to high values for a few minutes every hour can have the same number as just pain that stays the same throughout the time period.
==== Scales ====
[[Personal Science (book)]] recommends simpler scales to stop user from wasting time on details. I recommend somewhat wider scales that split positive and negative into 'definitely' and 'extremely/unusually' and the neutral state into two to let user guess which side is more likely.
Ratings drift is when user adjusts to some new average over time and that average becomes their new neutral. Sometimes this is caused by user not being to set in the ways they measure the phenomenon and exactly what ratings mean what. This is a problem because it creates false positives for tests and makes comparing between long time periods difficult. It is possible to look at distribution, especially extremes and guess at what happened.
I recommend anchoring a fairly wide scale to specific set of physical characteristics like how disruptive it was to user's every day work. This is less ambiguous and therefore faster too. Also this means comparisons can be done between different phenomenon that have the same scale.
{{Tool Queries}}
{{Tool Queries}}
[[Category:Tools]]
[[Category:Tools]]