The decisions behind the analyses.
Short, opinionated articles on the choices a researcher actually faces. CRD or RBD, AMMI or GGE, Method 2 or Method 4 of a Griffing diallel. Each piece walks through a worked example on the same data format StatVeda accepts, then sends you to the right tool to run it.
- Experimental Designs7 min read
When to use CRD vs RBD vs Latin Square
The three most common single-factor designs and how to pick between them. Same four-treatment dataset, three different ANOVAs, three different conclusions.
Read the article - Biplots, GxE9 min read
AMMI vs GGE biplot: which one and why
They look similar, they answer different questions. AMMI partitions GxE only; GGE partitions G plus GxE. Same 8 by 4 trial in both biplots.
Read the article - Plant Breeding8 min read
Diallel methods 1 to 4 (Griffing): which one fits your crosses?
Method 1 is the full set with reciprocals. Method 2 drops reciprocals. Method 3 drops parents. Method 4 keeps F1 only. Pick from what you actually planted.
Read the article - Experimental Designs7 min read
Augmented designs: when you have many candidates and few seeds
Federer 1956. When 100+ test entries each have enough seed for one plot only, replicated checks anchor the analysis. Fifteen test entries, three checks, two blocks worked through.
Read the article - Plant Breeding8 min read
Stability analysis: Eberhart-Russell vs Perkins-Jinks
bi (regression slope on environmental index) and s squared d (deviation from regression). When bi is greater than 1, less than 1, or close to 1. Same data, both methods.
Read the article - Experimental Designs9 min read
Split-Plot vs Strip-Plot vs Split-Split-Plot: which error structure fits your field
Two factors, but the randomisation restriction decides the error terms. Main-plot vs sub-plot precision, the strip-plot intersection error, the third stratum in split-split. Same factors, three different ANOVAs.
Read the article - Experimental Designs8 min read
Alpha Lattice vs Balanced Lattice: incomplete-block designs for large variety trials
Balanced lattice needs a perfect square and a fixed number of replications. Alpha lattice relaxes both. When 80 entries do not fit a clean square, the alpha design still recovers inter-block information.
Read the article - Plant Breeding8 min read
Line x Tester vs full Diallel: choosing a mating design for combining ability
A full diallel needs every parent crossed with every other. Line x Tester crosses a set of lines onto a few testers. Far fewer crosses, GCA and SCA still estimable. Pick from your crossing capacity.
Read the article - Diagnostics8 min read
Transforming agricultural data: when square-root, arcsine, or Box-Cox, and how Bartlett's test guides the choice
Counts want square-root. Percentages want arcsine. A skewed positive response wants Box-Cox. Bartlett's test on variance homogeneity tells you whether you needed a transform at all.
Read the article - Experimental Designs8 min read
LSD vs DMRT vs Tukey vs Scheffe: picking a mean-separation test without inflating error
Four mean-separation tests on the same treatment means, ordered from most liberal to most conservative. Why LSD over-declares, why Scheffe under-declares, and where DMRT and Tukey sit between.
Read the article
Want a worked example on your own data?
Read the case studies for end-to-end walk-throughs on realistic field datasets, then run your own data in the tool.