Summary of Data Errors and Fixes

 

 

 

Error

Fix

When fix is needed

 

sampling

replication

how much replication depends on the goal and on the population variation

bias

blind

human subjects, human observers of subjective things

random

virtually always when choosing things

human & technical

design of good protocols, use

‘knowns’  (proficiency tests)

 

 

 

always

RPA

better equipment

when greater accuracy is desired

Flawed analysis

Get independent analysis, publish raw data

Especially when data are processed heavily before use

all

good protocol

always