1 / 18

Layout of experiments & field trials

Layout of experiments & field trials. Eduardo Notivol C.I.T.A. Aragon P27 (Spain) Unidad de Recursos Forestales enotivol@aragon.es. Helmut Grotehusmann NW FORSTLICHE VERSUCHSANSTALT P7 (Germany) Abteilung Waldgenressourcen. Outline:. Design and analysis of experiments.

lynn
Télécharger la présentation

Layout of experiments & field trials

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Layout of experiments &field trials Eduardo Notivol C.I.T.A. Aragon P27 (Spain) Unidad de Recursos Forestales enotivol@aragon.es Helmut Grotehusmann NW FORSTLICHE VERSUCHSANSTALT P7 (Germany) Abteilung Waldgenressourcen

  2. Outline: Design and analysis of experiments • Basic Principles • Review of TBDX situation (questionnaire) • Working on Standards (Check – list) • Objectives (and duration) • Optimal or minimum req • Problems and limiting factors

  3. Basic principles: Experimental design Tool for addressing analytical problems without fixed laws 1.Replications. (experimental error basis) Standard Error of Difference Agronomic trails SED<1/3 diff Material selection SED<1/6 diff Knowing s2 & d ==> n 2. Treatment (broad sense = G.U.) Randomization 3. Local control of existing variation (Blocking and/or spatial allocation and analysis) Stat. Analysis (Models, Anova ...)

  4. Experimental design Initial hypotheses or constraints: Aditive Model Normality Homocedasticity. Different treatment errors are independient & distributed N(0,σ2) Statistic tests: N: Shapiro-Wilks, graphs distrib, freq acum., res * pred H: Barlett, Levenne, ratios variances Transformations No parametric methods

  5. Elementary Designs R.D. Model: yij = μ+ ti + εij Data.........; Proc GLM; Class treat; Model y = treat; Run; SAS

  6. Elementary Designs R.C.B Model: yijk = μ + ti +bj + εijk Data.........; Proc GLM; Class treat blq; Model y = treat blq; Run; SAS

  7. F E.M.S. ¿Fixed o Random? • Critic decision • Not well documented on texts • Usually based on subjective statistic agreements Fixed: Levels of factor clearly targeted or selected Results & conclusions from anova are for these levels Main aim: Mean estimation of the variable for each level (BLUE) Random: Levels are a random sample from all posible. Results & conclusions from anova can be extrapolated + levels Main aim: Variability estimation of the variable or factor or perhaps prediction at a given level (BLUP)

  8. How to asses? A PRIORI ¿Fixed o Random? Scientific Criteria : 1) is it possible to repeat the factor levels in other site or year? 2) has it meaning this replication? Yes + Yes = Fixed Statistic Criteria : “Random” few levels (3-5) =>weak variance estimation, Better setting as fixed and use the results only at these levels “Fixed” with many levels (>10) without structure, better setting as random and estimating means by BLUPs E.M.S. Numeric difficulty SAS

  9. σ2e+ c1Φα σ2e+ n σ2ab +nb σ2a

  10. Incomplete Blocks Evaluation : high nº genotypes limited material ‘Many genotypes’ means huge blocks # no control I.B. Not all treat by block, so several blocks are needed for a complete replication Based on Aditivity: B-C= (B-C)3 = (B-A)1-(C-A)2 Experimental error independent of treatment

  11. Incomplete Blocks • Coexist direct & indirect comparisons • Lost of accuracy on indirect comparisons • but experimental error reduction Resolvable designs i.e.: g=k bi α-latice, latinized, row-columns,... Complex specific Software interblock info

  12. I.B. design Efficiency Objetive: To compare genotypes (at max. accuracy) E = (SEDRCB/ SEDIB)2 nº of extra replications in a RCB to get same accuracy level An IB with 4 reps y E=1.5 equals to a RCB with 4x1.5=6 CB “Efficiency” ~ costs

  13. RCB Conf. Int. RCB IB Conf. Int. IB

  14. 28 genotypes 4 blocks of 7 Latinized Row-column Row-Column 3 rep -lattice

  15. Results from a Questionnaire • Kind of designs • Duration • Technical aspects • Some statistical issues (number of “r”) • In general only few parallel sites • •More replication/site than sites/series (i.e. parallel sites) • Demand for adjusting layout for joined breeding activities (Plant/plot, spacing, duration of testing) • •Different designs in the past, designs for the future?

  16. Design= f (objective) • Evaluation of variation (prov, prov-prog) • Approval of tested MFR (prog, so, cl) • Selection of parents for breeding (prog, cl)

  17. MISSION: STANDARDSOPTIMAL MINIMUM REQUIREMENTSCHECK LIST

More Related