教授有版权，所以看了就好。 ---------------有情有义分隔线------------------------ 以下是教授的回答： Just pick the version which makes the test statistic greater than one. We speculate the residuals follow certain rules, such as bigger the variable "room", bigger the residuals. However, we are just trying to find a pattern, if it goes the other way, say smaller the variable "room", bigger the residuals, it's still a pattern of heteroskedasticity. It doesn't matter which one is the case. As long as either RSS2/RSS1 or RSS1/RSS2 is significant, we conclude the existence of heteroskedasticity. If you pick the version with RSS2/RSS1 < 1 (by default RSS1/RSS2 must be greater than 1), then there's no chance your test stat can be significant and you are not doing the right test. We just want to show these two RSSs are different, it really doesn't matter which one is bigger. ----------------------------------------------------------------- 大家加油～～ --

※ 发信站: 批踢踢实业坊(ptt.cc) ◆ From: 203.67.39.70