 |
 |
If trying to locate a lot of projects relative to one another, a possible approach is to formulate a set of notionally independent questions to ask of each one. E.g. How great is the precision? (Whatever that means.) How great is the integrity? (Whatever that means.) How great is the completeness? (Whatever that means.) How great is the funding? (No doubt what this means.) How many hits on the associated web site per month? (No doubt here, either.) ... The responses would have to be quantified some way, perhaps by suitably formulated questions in conjunction with a Likert scale: http://en.wikipedia.org/wiki/Likert_scale There would be a level of subjectivity. Nevertheless, if the questions were well chosen and the assessment done in a consistent way with checks and balances in place, the information thereby gleaned may be valuable. Once the survey results are available, it is reasonably straightforward to use a multivariate technique such as multidimensional scaling (MDS) to produce a visualisation of similarities among the assessed projects. More on MDS here: http://en.wikipedia.org/wiki/Multidimensional_scaling The R statistical environment is a great help when it comes to analysis: http://www.r-project.org/ Best, Tim Finney On Tue, 2010-08-31 at 00:00 +0100, Leif Isaksen wrote: > Hi Robert > > I think there's certainly some merit to such an approach. Along with > precision however, you'll need to take account of accuracy. I'm not > sure that lower granularity would go hand in hand with less integrity > though. Generally speaking, as precision increases it requires greater > effort (and in terms of derived data, becomes impossible) to maintain > the same level of accuracy (and as one generalizes so it becomes > easier). > > A further axis to consider would be completeness - to what extent is > the level of semantic information maintained - but this operates at > different levels: a lacuna in the text may remove important > information, and equally a gap in our knowledge may do the same (so a > Linear A tablet can be physically complete yet inaccessible to us). > Presumably there is a commensurate axis for (unidentified) inclusions > as well but I don't know what the word for it would be :-) >