This repository contains a pair of Lua scripts to approximate curves in the form t(n)=xn^k and t(n)=10^xn^k*log(n), which can be handy when using performance metrics to gauge the complexity of a procedure.
$ lua -e "function f(n) return 3.57*n^1.93 end for i=1,100 do print(i,f(i)) end" > /tmp/x
$ lua nlog.lua /tmp/x
t(n)=10^1*n^1.196*log(n)/log(1.6225644721121)
with precision avg=73.672%, stddev=213.080%
$ lua n.lua /tmp/x
t(n)=3.5700000000015n^1.9299999999999
with precision avg=100.000%, stddev=0.000%
$ lua -e "function f(n) return 10^-3*n^1.23*math.log(n,1.44) end for i=100,1000 do print(i,f(i)) end" > /tmp/x
t(n)=0.0056364978850545n^1.406
with precision avg=99.097%, stddev=16.923%
$ lua nlog.lua /tmp/x
t(n)=10^-3*n^1.23*log(n)/log(1.4399999999999)
with precision avg=100.000%, stddev=0.000%
Customary statistical precautions apply. You will most likely want a relatively large sample size; if your dataset consists of two points, your best approximation will be a straight line! It is also important to distribute these points over a wide margin. Any continuous curve sampled at a sufficiently short interval will look invariably flat.
This repository is in the public domain.