Quote:
Originally Posted by pwalker8
But they don't point at any authors, that's the issue. This is a model without any validation. It's not a sampling of 10,000 authors, it's some guy who who tries to figure out what the actual sales numbers are based on various indicators. If he was actually comparing his guesses to the real sales figures for a sampling of 10,000 authors that might be a reasonable way to validate his model, depending on the sampling. You can't calculate something from nothing even with basic statistics. Garbage in/Garbage out.
|
Except that they do use actual data from actual authors, at least according to their methodology and everything I have seen. They don't share who those authors are and they don't share their calibration data (due to privacy issues) but they do have actual data. And need it in order to make sure it works.
At the original link:
Quote:
|
All of it calibrated against 700,000 points of raw, unfiltered daily sales data, from over 20,000 distinct ebook titles across all 15 stores."
|
It may not be quite 10,000 authors but it is 20,000 books, so it is at least a decent size.