.ASEALLY (continued) on how peskedyou think that this map locks Like ahead off tine and one of the fimctions of a real pilot program would bel te find out precisely this, Any sampling design has te prods really twothings. One, anostinate of the quantity you after and also en estimte of itsintrinsic variehility. So | it turns out if this thing is very peaked, the way to go this problem is to pick a few pointe initially at random, and then sample around it in big clusters. This wuld man & you take many samples near this point, if the thing is If it ts quite macoth, then one spreads his samples Py peaked. arowl so that this gives you the optinal estimate for a fixed ancynt of money. Where it's wrnlly the case that if one amples in this vay, around a certain location, it costs less per +t So that scne guesses are going to have to be nade. Cne | to get dova to List of Just what it fe that we vant to simpleme us. ve can't really just be satisfied vith Just saying Nev Paglands | You have to talk shout vist it is you are going to do! | tate particular aren. Mow in thinking abyut this probley 1 nb cody” portion of it that I've locked into fnany wy is the in . aempling proble. lanmiig setting caite ali these prothe of getting the actual material, I think that ve ought ta aim at getting a correlation betveen something caloulated from abate data, for particular aress. What I vould have in mind due to the fact thet the food distritution in this country 16 90 wHespread that one would try to take weighted average of say a cincle _ of — roughly 150 miles around a city vhich generally represedtsite we . aoe ee cadat i te . / of