pro mc_eror, g, sims, bndu, bndl, confidence=confidence, alas=alas, _extra = e ;+ ;procedure mc_eror ; ; simple procedure that calculates confidence bounds ; about a given parameter value g given a certain number ; of monte carlo simulations contained in array sims. ; assumes the simulations are distributed normally about ; the 'candidate' parameters. ; ; by default,set to find smallest possible bounds. can also ; be set to find confidence bounds by counting upward ; and downwards by required fraction. ; ;syntax ; mc_eror, g, sims, erru, errl, confidence=confidence ; ;parameters ; ; g [input, required] parameter about which to estimate bounds ; sims [input, required] array of monte carlo simulated g's ; minimum 10 simulations [though this ; seems a ridiculous number] ; bndu [output] upper confidence bound ; bndl [output] lower confidence bound ; ;keywords ; ; confidence [default = 0.68269] confidence limit at which bounds ; are desired. ; alas find confidence bounds by counting upward ; and downwards by required ; fraction. may also be set ; to find smallest possible bounds. ; ;history ; LL Jan03 ;- LL Apr03 add smallest interval mechanism and turn old into ; frac keyword ; ok='ok' & ng=n_elements(g) & nmc=n_elements(sims) if ng eq 0 then ok='No g defined' else \$ if nmc lt 1 then ok='Please, more simulations.' if ok ne 'ok' then begin print,'Usage:mc_eror, g, sims, mcup, mcdown, confidence=confidence' if ok ne 'ok' then message,ok,/info return endif if keyword_set(confidence) then begin if (confidence gt 1.0) then begin ok= 'Cannot be more than 100 percent confident unless W says so. ugh.' message, ok, /info return endif conf=confidence/2 endif else begin conf=0.68269/2 endelse szsim = size(sims) if (szsim(0) gt 1) and (szsim(1) eq 1) then sims = transpose(sims) if keyword_set(alas) then begin auxar = [g, sims] & argg = auxar(sort(auxar)) & a0w = where(argg eq g) a0w = median(a0w) & intt = round(nmc*conf) & nmcc = nmc-1 ;IDL index corr. if (a0w + intt le nmcc) and (a0w - intt ge 0) then begin erru = argg(a0w+intt) & errl = argg(a0w-intt) endif if (a0w + intt ge nmcc) then begin erru = argg(nmcc) & errl = argg(a0w-intt-(intt-(nmcc-a0w))) endif if (a0w - intt le 0) then begin errl = argg(0) & erru = argg(2*intt) endif endif else begin auxar = [g, sims] & argg = auxar(sort(auxar)) & a0w = where(argg eq g) a0w = median(a0w) & intt = round(nmc*conf) & nmcc = nmc;-1(inx corr)+1(g added) btbd = a0w - 2*intt > 0 ;lowest possible lower bound in IDL index space tpbd = a0w + 2*intt < nmcc ;highest possible high bound in IDL index space testint = fltarr((tpbd-btbd-2*intt)+1) ;tricky.... for j = btbd, tpbd-2*intt do testint(j-btbd) = argg(j+2*intt)-argg(j);calculate width of each interval jj = where(testint eq min(testint)); identify smallest width ; and corresponding bounds bndu = argg(jj+2*intt+btbd) & bndl = argg(jj+btbd) endelse ;for j = 0, a0w do testint(j) = (auxar(j+2*intt)-auxar(j)) ; ; ; if (a0w - intt lt 0) then begin ; testint = fltarr(a0W) ; for j = 0, a0w do testint(j) = (auxar(j+2*intt)-auxar(j)) ; jj = where(testint eq min(testint)) ; erru = auxar(jj+2*intt) & errl = auxar(jj) ; endif ; if (a0w + intt gt nmcc) then begin ; testint = fltarr(nmcc-a0w) ; for j = a0w, nmcc do begin ; testint(j-a0w) = (auxar(j)- auxar(j-2*intt)) ; endfor ; jj = where(testint eq min(testint)) ; erru = auxar(aow+jj) & errl = auxar(a0w+jj-2*intt) ; endif ; if (a0w + intt le nmcc) and (a0w - intt ge 0) then begin ; testint = fltarr(2*intt) ; for j = a0w-intt, a0w+intt do begin ; testint(j) = (auxar(i)- auxar(i-2*intt)) ; endfor ; jj = where(testint eq min(testint)) ; ii = nmcc - jj ; erru = auxar(ii) & errl = auxar(ii - 2*intt) ; endif end