Quote:
Originally Posted by JustinD
I perfectly trust that you have more statistical skill than myself and I did like to hear the defense of the variables that most would not take into account as in my mind the addition or subtraction of these variables is why statistics are generally garbage. Is there a greater chance of a gun injury if one exists in the home… well no crap, will there be more stair falls in a multistory home versus a single level ranch home? Yes! Wow, that was a useful study.
My question lies on this, after listing a plethora of variables needing analysis you state “a few hundred sales is more than sufficient”. Seems it would not be as that group of a few hundred may contain single examples of many of those items, two, or be excluded completely as it is being culled from many areas used by new or less informed collectors such as PWCC and eBay and has no access to private sales.
The idea that this is a statistical fact requires a large leap of faith and should not cause some stone wall statements of hyperbole touting the need for this. I’ll let you enjoy the service, my personal opinion is that the cards you used as examples were well graded based on the chipping and corner wear. Yes, some other areas were exceeding but the grades are a sum of the parts. If the sticker makes you pleased and happier as a collector then please do.
I consider these services snipe hunts and it’s not changing. I do not believe in a hobby that has no real price stability and changes daily that a market of constant end value flux can be measurable without faith.
So after all that time wasted in my ramblings… do what you want for yourself but please understand the sales pitch is not going to have any true impact. You are a corvette salesman pulling over an Amish buggy to make a sale. It’s not going to be easy.
|
Not sure why this point isn't getting through, so I guess I'll just say it again? These are for me. I'm not sending these cards in to flip or make money from the stickers or anything like that. I just like them.
As far as your concerns about the limitations of modeling goes, you would need to have a sufficient number of examples of cards for each variable you're instead in evaluating in the model. So if you want to understand the impact that something like tape stains has on sale prices, you'll need to have a sufficient number of cards with tape stains in your data set. How many cards that number needs to be in order to be "sufficient" depends on how large of an effect it has. If that feature is something that nearly doubles the value of a card, then you might only need like 20 or 30 examples in your dataset that have this quality. However, if it's something that only adds or subtracts say 5% to the value of a card, then you'll need a lot more examples in your dataset in order to demonstrate that the effect is statistically significant. If you only have a few cards with tape stains in your dataset, then you won't be able to say anything about its impact regardless of how many total cards are in the model. Also, you should probably just exclude those few cards from your dataset if that's the case because you'd be introducing unnecessary variance to the model, and the whole point of building a model to begin with is to explain the variance in the sale prices.