Solutions to exercises May 8-10, 2006 Iowa State University, Ames – USA Jean-Louis Laffont Kirk Remund
ISTA Statistics Committee2 Qualitative testing plans – Exercise 1 “ I strictly do not want to have any GM seeds in my conventional seed lots ” 1. Is this possible to check? If yes how?, if no why? 2. Try to find ad hoc testing plans with Seedcalc LQL=1% beta=approx. 5% LQL=1% beta=approx. 20% LQL=1 per 1000 beta=approx. 5% LQL=1 per 1000 beta=approx. 20% LQL=1 per beta=approx. 5% LQL=1 per beta=approx. 20% 3. Choose 2 LQL values, ie levels you wish to reject often and two beta risks (how often you reject this level) => 4 different cases Try to find ad hoc testing plans with Seedcalc
ISTA Statistics Committee3 We would have to check each individual seed Color, or specific fluorescence introduced on purpose, … able to see easily and cheap … If we look at a few seeds, –we can know, there is presence, if we find some GM seeds, –but we can not tell “ there are no GM Seeds ” if we did not find any on few seeds We can look until we find one, …. and reject; or find none =>We must look at as much seeds as needed to have a “ good idea ”, Option1 = we wish to rarely accept a lot with a given % of presence, Accept=0 LQL= this Low Quality Level Beta= how often we will accept a lot which in fact had this LQL Option2 = we define how many seeds we can look at (time, money … ) and see the corresponding risk Qualitative testing plans – Exercise 1 1. is this possible to check? If yes how?, if no why?
ISTA Statistics Committee4 Find Plan with accept=0,LQL=1%, beta=5% 2 pools of 150 seeds A possible answer Qualitative testing plans – Exercise 1 2. Try to find ad hoc testing plans with Seedcalc
ISTA Statistics Committee5 1) Type LQL 1% 2) Select a convenient number of seeds per pool 3) Click find plan with zero tolerance How to get an answer:
ISTA Statistics Committee6 keep LQL 1%, but accept a greater beta risk, ie 20% instead of 5% For instance, change seeds per pool (from 150 going down) until beta=20% approx 2 pools of 80 seeds
ISTA Statistics Committee7 Find Plan with accept=0, LQL=0.1%, beta=5% 10 pools of 300 seeds 3 pools of 1000 seeds 1 pool of 3000 seeds
ISTA Statistics Committee8 keep LQL 0,1%, but accept a greater beta risk ie 20% instead of 5% For instance, keep 10 pools and decrease nb of seeds per pool 10 pools of 160 seeds (instead of 300 seeds for beta= 5%)
ISTA Statistics Committee9 Find Plan with accept=0, LQL=0.01%, beta=5% 100 pools of 300 seeds
ISTA Statistics Committee10 keep LQL 0,01%, but accept a greater beta risk ie 20% instead of 5% For instance, choose 50 pools as a maximum 50 pools of 320 seeds 1)50 pools 2) Change # of seeds per pool, until beta is approx 20%
ISTA Statistics Committee11 3. Choose 2 LQL values, ie levels you wish to reject often and two beta risks (how often you reject this level) => 4 cases try to find ad hoc testing plans with seedcalc Your aimAn appropriate answer: LQL1= beta1= =>nb pools: nb seeds per pool: LQL1= beta2= =>nb pools: nb seeds per pool: LQL2= beta1= =>nb pools: nb seeds per pool: LQL2= beta2= =>nb pools: nb seeds per pool: Zero tolerance
ISTA Statistics Committee12 1. “ Xyz ” is ready to pay for passing seed lots tested to the following criteria: AQL 0.3% at alpha=5% and LQL 1.5% at beta=5%, what shall I do? Qualitative testing plans – Exercise 2 Pool size set up to 100
ISTA Statistics Committee13 2. Find different possibilities for the following objective: AQL= 0.5% LQL=1% alpha=5% beta=5% Qualitative testing plans – Exercise 2 Accept if # of deviants does not exceed 300 seeds/pool 39 pools seeds seeds seeds seeds seeds
ISTA Statistics Committee14 Quantitative testing plans 1. Is the testing plan appropriate? NO
ISTA Statistics Committee15 Quantitative testing plans
ISTA Statistics Committee16 Getting an estimate of % of GM - Exercise 1
ISTA Statistics Committee17 Getting an estimate of % of GM - Exercise 2
ISTA Statistics Committee18 Getting an estimate of % of GM - Exercise 3
ISTA Statistics Committee19 Getting an estimate of % of GM - Exercise 3