Testing statistical significance - Revisiting.

[Clariman]

I would like to resurrect this question now that I have done my research and written most of it up (perhaps late to return to statistical significance). First a brief summary of what I was researching and how its focus has shifted a bit, then a brief summary of my results, following by a re-framing of the stats question. Original questions and responses here posting.php?mode=reply&f=9&t=14959

My Research
I have analysed 320 artefacts dating from around 300 AD to 900 AD that have a shared vocabulary of symbols that are unique to this culture. I have recorded over 1,000 appearances of the designs on these artefacts that include; pre-Christian monoliths, Christian cross-slabs, personal jewellery, cave-drawings and gaming pieces. It has often been suggested that the symbols appear in pairs but it has been acknowledged that they don't always do so - some appear singly, others appear in slightly larger groups. I have created a hypothesis that defines 3 different types of Symbols and have documented how the types differ, in terms of their designs, syntax and, therefore, probably purpose or meaning. I have analysed every appearance of every symbol in order to design and test my hypothesis.

Results
The symbols I have defined in Group 1 conform to my hypothesis 93 to 95% of the time. The ones in Group 2 about 95%, Group 3 100%. Looking from an artefact perspective about 95% conform to my proposed Syntax, using these 3 different types of symbol. I have analysed the exceptions and many can be explained but cannot be proven.

How do I measure statistical significance
To my numerate (but not mathematician) head, I am happy with how well my hypothesis fits. It may not be 100% but it isn't far away and no-one else has measured and tested it in this way, so I'm happy with the results. However, I'd like to double-check that I am not overstating it. How can I test this more rigorously with proper tools? Note that I have excluded the geographical aspects that I had previously talked about.

Many thanks
C

[csearle]

I have available a brilliant statistician in Muenchen, to whom I have forwarded your post. I hope he has the time and inclination to answer.

Chris