Here is my data for 20" New Stamps. You need to focus on just New Stamps because we believe that the prices for Old Stamps tend to be higher than New Stamps and this factor is large enough that you need to take stamp era into account. I'm not yet in a position to say whether Intermediate expected values are different from New Stamps or not (based on statistical analysis taking into account condition, weight, etc) because I haven't done the statistical analysis. I haven't done it for Old Stamps either, but I believe that difference will be significant in analysis even though I haven't done the model fitting yet.
This is the latest data:
[img]http://black.net.nz/cym2015/k-new-20-part1.jpg[/img]
And this is earlier data which I haven't yet moved/recoded over into the new layout:
[img]http://black.net.nz/cym2015/k-new-20-part2.jpg[/img]
Note that I'm dropping the Newer/Older New Stamp distinction in my current data coding. I made that decision to get larger samples even before questions about the weight of evidence supporting such a division.
Under the Bids field if it says -3 that means it was either sold for an undisclosed amount or the auction was removed, so it is a flag that the record doesn't contribute good price data. If Bids is 0 then nobody was willing to bid at that price. In both cases the Price field records the asking price. I use other codes (all documented in the coding tab of the spreadsheet of course). All my data is freely available to anybody who wants it -- but still rather messy at the moment, as you can see.
So what do I see eyeballing the data? An expected range of $900 - $1400 say my eyeballs. By expected range I mean expect maybe 80% of sales to fall in this range, 20% to fall outside this range. This is sort of equivalent to "Price is ok for buyer and seller. Not a steal by the buyer, not a rip-off by the seller." But that's an individual interpretation and others will have different ideas. Which is why I like to present the raw data.
More accurate than that means doing the statistical analysis properly, and that needs bigger sample sizes. In addition to considering condition and weight there are also factors like seller reputation (eg Hazelshould) and quality of ad/soundfile/pics. It is also likely that being in the USA/Canada vs Europe is an expected price factor, and I haven't yet even got a handle on that with the massive 602/Sound Creation database. Better economists/statisticians than me have pondered and analyzed such things and I know it isn't easy. The mathematics is ok, but getting quality data? Tough stuff.