Pink noise about literaturame

[JJ]

I'll check this out. The fitting code is in the paper itself, from line 624:

    spec.fit <- lm(log(this.spectrum$spec) ~ log(this.spectrum$freq))

So it's a loglog fit of coefficients. I'll try to do it with squares. Which is actually just 2*log, right?

[ccottap]

Something odd going on here. The coefficients had not changed so I took a look at the code and saw something misplaced: the fit should be to the log of the aquare, rather than to the square of the log, thus:
spec.fit <- lm(log(this.spectrum$spec) ~ log(this.spectrum$freq^2))
Now, this may be a trivial issue but it is evading me: by doing so, the coefficients are actually halved rather than doubled. I do not know R enough to debug this.

[JJ]

I already did that:

 spec.fit <- lm(log(this.spectrum$spec) ~ log((this.spectrum$freq))**2)

It's actually the same. What is being fit is to the left of the ~ so it's natural that it goes that way, no big deal.

[ccottap]

log((this.spectrum$freq))**2 is the square of the log, and it should be the log of the square: log((this.spectrum$freq)**2). But there is another issue: check the coefficients of the raw log-log fit and those of the log-log(square) fit. The latter should be twice the former, but it is not the case (unless there is a problem on my side when compiling the document).

[JJ]

OK, will look into that.