Whilst conducting research for a question I asked in Psychology.SE I came across a report which uses parsimonious statistical comparisons and although this Stats.SE question has an accepted answer, it doesn't really explain to me sufficiently what parsimonious statistical comparisons are and how they make a difference compared to non-parsimonious statistical comparisons.
The report is an evaluation on the effectiveness of SOTPs (Sexual Offender Treatment Programs) and the report states in Annex C, that:
A wide range of potential matching factors were identified following extensive review of the literature and in consultation with colleagues...
Many of the factors are categorical and so were converted to ‘dummy variables’. The final model, referred to as the 'parsimonious' model, included factors that were either deemed theoretically important (asterisked in Table A.5), and/or were empirically related to both selection onto the Programme and one of the main five outcome measures
Table A.5 (pages 39—43 of the report) is a bit big for posting here, but here is an example of parsimonious and "less parsimonious" results presented in the report.
I know about different kinds of average - Mean, Mode and Median, but anything else to do with statistics can be above my understanding, so I am a complete novice with regards to examining statistical data. I tried looking at this explanation on StatisticsHowTo.com and I am still none the wiser.
Maybe I am a dunce in this field, but can someone please explain in layman's terms
- What are Parsimonious Statistical Comparisons and how did the parsimonious model affect the statistical data? and
- Would the parsimonious model have made the results of the comparisons more accurate?