-jinx- said:Maybe I made a mistake by typing so much, since you seem to have missed my point. I think this phrase from your response sums it up: "....the data is so inaccurate in the first place and the aim of it is to show basic trends."
Since you admit that the data is of completely unknown accuracy and you're making assumptions with no more justification than saying they are "reasonable" (which is circular), why should anyone believe your conclusions? Your "proof" is completely unconvincing for those reasons.
Well maybe you're just trying too hard.
If I had scales that said an apple was 90g and another was 210kg then I think I can reasonably suggest that the second one is heavier, even if we have no idea of the accuracy of the scales. Even if was only to the nearest 20g it's gonna make no difference to the overall argument.
You're just being anal really. The numbers show discrepancies of 3-4m- 8m vs 4m for example. This is a hell of a discrepancy and you're gonna struggle to argue it by going down the road of "you don't know the accuracy of these numbers". It wouldn't matter if the 4m number was 30% out, it's still significantly less than the 8m number being claimed.
It is a very reasonable assumption that the margin of error is pretty consistent from 6 month period to 6 month period. We have to account for fluctuations of course, which is where the simplification is coming in, but why would the error be anything else? We are talking about random noise here. I have a number of reasons why I think this is true:
1) The only time the two lines above become close is when we know there are shortages. Bit of a coincidence if the numbers are being so wildly calculated.
2) If the average of ten numbers is 1.2, this can be represented as 1.2 with some random fluctuation imposed on top for each month. Some months it's higher, others it is lower. I guess you could argue that we don't know exactly how much higher or lower each month which is fair and this is the only bit that I could agree is uncertain, although common sense would suggest that there is no reason why it should fluctuate that much. NPD has actual sales for over 60% of the market and extrapolate up for the other 40%. Over a 6 month period it would be reasonable to assume that they have overestimated as much as underestimated with the means matched surely? It's a long enough sample time for the fluctuations have been damped down.
3) The pattern as can be clearly seen in the graph and explained a number of times by me. Let's think about the number we are looking at here. We KNOW that sales from Sep-Mar over the xmas period are around double that from Mar-Sep the quiet period. This goes totally against the pattern of shipments made from Sony who ship more in the summer / autumn to prepare for xmas. The only way the figures could be fudged so that shipments = sell through would be to assume that NPD is only reporting about 1/3 of sales during the summer and is about 2 times to large over xmas which would obviously be absurd !!!
A bit of common sense goes a long way here...