The curves for compile time and binary size are double exponential, $size=e^{e^{a*Level}}$, where a is 0.79 for compile time and 0.12 for binary size. Not sure why the author thought a single exponential looked like a good fit.
R code:
csh=read.csv("c:/web/csh.csv") plot(csh$Level, log(csh$Compile_Time), log="y") csh$ll_CT=log(log(csh$Compile_Time)) c_mod=glm(ll_CT ~ Level, data=csh) summary(c_mod) csh$ll_BS=log(log(csh$Binary_Size)) b_mod=glm(ll_BS ~ Level, data=csh) summary(b_mod)
Data extract from table:
Level,Compile_Time,Working_set,Binary_Size 5,1.15,54288,135680 6,1.22,59500,788992 7,2.00,70728,4707840 8,6.43,121852,28222464 9,33.23,405472,169310208 10,202.10,2141272,NA
The curves for compile time and binary size are double exponential, $size=e^{e^{a*Level}}$, where a is 0.79 for compile time and 0.12 for binary size. Not sure why the author thought a single exponential looked like a good fit.
R code:
Data extract from table: