TheSteve said:
uber said:
i'm not a business expert, but i am a mathematics expert, and it is highly ignorant to use a linear regression model to say anything at all about our present discussion. yes, the ps3 is selling just over 100k a week, but that does not mean you can treat that number as a mean. price cuts, game releases, and holidays all cause spikes in the numbers. the only meaningful way to talk about the numbers is to speak as generally as possible over the whole data set. the ps3 has clocked in about 10 million units sold a year. if we were to assume linearity 2009 would yield about half that. but that's not the case. if you really wanted to predict performance one would have to look at several key points in the year and draw a percentage. this could allow one to extrapolate the rest of the year. |
By that same logic, Killzone should sell 166,000 copies next week, because it's been out for 9 weeks and sold 1.5 million, thus they must be selling 166k a week. That ignores trends, diminished utility, and actual data we have right at our fingertips... Yeah, there was a spike with the Japanese release, but Killzone will likely be selling less than Halo 3 again next week, not rocketing into 3rd. That linear regression is based on ACTUAL sales data, not on the current numbers extrapolated. The PS3 is the only console this generation to have a negative slope in its sales... It's selling FEWER concoles year-over-year, indicating that the gap will actually increase the longer both consoles are on the market.
PS3 moves fewer consoles each year: fact. If they continued selling at the CURRENT rate they're moving (which would actually buck their current downward trend) it would take nearly two years to close the gap with their nearest competitor (if said competitor wasn't selling anything in that time).
100k may not be a good mean to use, but it's actually being generous given the trend. However, you using the average given the obvious spike last year and the drop since is misleading. Median would actually be the better indicator of central tendency due to this very fact.
I would think you'd know this, being a statistician and all.
*bumps knux*
|
*does double take* am i reading that right? you respond to my post condemning linear regression and then attack my argument as if it were in favor of it? are you on drugs? games sales tend to follow more of an exponential curve than linear. your give a pathetic straw man argument using killzone as an example and then flaunt your statistical prowess like you are john freakin nash? impressive. if you actually paid attention to my post you would have seen a perfectly functional way to extrapolate sales. to extrapolate does not mean to regress linearly, which is what you seem to think. whether you talk about a mean or median you are still trying to talk about slopes and what not. well anyone worth his salt knows linear models are shitty at describing non-linear phenomena.
it would be silly to model sales based on history (specifically in the case of the ps3) if key points are not roughly the same (such as the same month in a previous year). it's fine to say that the trend is negative, but as soon as you want to talk about slopes, then you get absurd statements like sony selling a negative number of ps3's in x years. that is a little remedial for me, which is what i was trying to explain to you.
and i am not a statistician, i am a mathematician. there is a difference. you didn't know, so i won't make a stink. but in the future, we don't like being called statisticians. i guess it's like calling a stripper a prostitute. there is a good deal of similarity, but boy do they get pissed when you do it.
