This depends upon seed available with you. If you do not want to go for augmented then conventional will work.

We have sufficient seeds. Please suggest the name of conventional designs?

Handling a large number of treatments may result in inefficiency. Can you divide it into batches and use RCBD, please?

I suggest a randomized complete block design (RCBD), but the number of genoypes is too high, it will be necessary to do this study in stages. However, I am willing to collaborate with you. You could write me at this mail: [email protected]

A conventional design is a randomized block design, where the experimental units are in groups called blocks. The treatments are randomly allocated to the experimental units inside each block. When all treatments appear at least once in each block, then there is a completely randomized block design.

Dear @Andrew Paul McKenzie Pegman,

Thanks for answering. I knew about RCBD. But would growing 120 Genotypes in RCBD is statistically sound or not?

Dear researcher, greetings

You could use the square lattice design that is recommended for a large number of treatments, in your case 120 genotypes. However, there is a major limitation that would be the loss of some treatment, as it would compromise the entire mathematical structure. Thus, I recommend using the randomized block design, as it is a robust design and you will not have problems in case of loss of treatments or even entire plots.

Since the number of entries is very high, it is important to use an incomplete block design. I think alpha-lattice (0,1) is the most appropriate design

To Parmeshwar K. Sahu, the kind of work you suggest is important and must be done, but I am not certain that any statistical design is going to handle such huge numbers of genotypes. Below I attempt to conform to prevailing practice, but I have mixed feelings.

Attempting to screen large numbers and getting the statistical results to mean a lot is a subject that would make for a dozen major doctoral dissertations. What if you get significance at alpha = 5%? That does not tell you which is better than which. Multiple comparisons of high numbers of genotypes would require enormous numbers of replications. Then there is the problem of power, custom demands at least 80% for research and that too adds to cost in terms of replication rates.

For any particular characteristic (yield, or insect resistance, or disease resistance, or ...) a good statistician might suggest you look at your 120 genotypes and choose 5 for each characteristic. Then for each characteristic run a separate experiment (for instance a complete block) using the 5 chosen genotypes plus a popular control and requiring 5% significance to justify moving on to multiple comparisons. Then you can make each of the 5 comparisons with their control (six genotypes counting the control) seeking a 1% significance each comparison. The control might be the best variety for that characteristic that is now in use by farmers where you are.

Then instead of a frustrating huge experiment you have to grow out all 120 only once to select some seemingly better genotypes and then you can do several practical small trials. (If you can get an idea of the standard deviation of rice data in your region, it might help you consider the sample sizes, number of plots, you really need, maybe summing the requirements for all the comparisons in each of the small groups.) Some authors suggest you minimally design to have at least 20 df for each omnibus trial (such as a RCB ANOVA) however even that is unlikely to be enough. (Beware of magic numbers called 'small'. 'medium', and 'large' they are deprecated in literature and, tacitly, by their originator too.)

Best wishes.

You could perform a stability analysis also known as a Genotype by Environment analysis (G*E). A G*E analysis is nothing else but a Factorial Design in which you're submitting into analysis the performance of a group of germplasm for a particular set of traits. In order to do this, you need to test your cultivars/hybrids in multiple locations across multiple years (no less than 3) or growing seasons. In some countries, by virtue of their environmental conditions, researchers can perform between 2 or even 3 evaluations per year especially if the evaluations are performed in the confinement of experimental stations.

Generally speaking, you would use the same Randomized Complete Block Design (RCBD) with no less than 3 replication per treatment (genotypes) in every location each year/growing season & in the process collect all the data of interest [e.g., plant height, days to anthesis, days to harvest, genetic gain (grain or biomass yield), etc.].

By the time you finish collecting the data which you should save it in an MS Excell sheet then you can proceed to perform the Analysis of Variance (ANOVA). I suggest you do the analysis using RStudio since it is an open-access license.

Depending on the number of factors involved in the analysis you should have a statistical model similar to this:

yijk = µ + αi + βj + γk + δl + (αβ)ij + (αγ)ik + (βγ)jk + (αβγ)ijk + εijk

Where:

µ: intercept or the overall mean of the population.

αi: effect of the Genotypes (j: rice hybrid 1, rice hybrid 2, etc.).

βj: effect of the blocks (effect of the distribution of treatment's replications in blocks).

γk: effect of years/growing season (k: year 1, year 2, year 3, etc.).

δl: effect of locations (l: location 1, location 2, location 3, etc.).

(αγ)ik: effect of the double interaction between Genotypes & years/growing season.

(αδ)il: effect of the double interaction between Genotypes & locations.

(γδ)kl: effect of the double interaction between years & locations.

(αβγ)ijk: effect of the triple interaction between Genotypes, years, & locations.

εijk: effect of the random experimental error which should satisfy the following conditions: ɛ: ~ N (0, σε2).

Additionally, keep in mind that each collected data set (one per location per year) should satisfy the following conditions:

1.- Normality of residuals (experimental errors).

2.- Independence of residuals (experimental errors).

3.- Homogeneity of variances.

The reason I always recommend performing a G*E analysis instead of a simple Completly Randomized Block Desing (CRBD) is that you can measure the performance of a given set of genotypes (in your case rice varieties/hybrids) across multiple environments for a particular set of phenotypical and physiological traits which otherwise would be measure only in one particular environment which of course represents a limitation when you try to extrapolate your results to other locations where the set of genotypes were not tested.

Finally, when you finish running the analysis you should perform the interpretation of the results starting from the triple interaction (Genotype*Environment), understanding that, in statistics, the interaction between locations & years is considered as "environments", in other words, Genotype*Location*Years = Genotype*Environment. Then, you move towards the double interactions & finally to the mains effects (i.e., Genotypes, years, locations).

Although, this is a very succinct response to your query I hope it helps.

I think because the number of entries (120 genotypes) is very high, the incomplete block design is useful. The alpha lattice (0,1) design is the most appropriate.

Using RCBD for more than 20 genotypes leads to very big blocks in such case there is likely to be great soil heterogeneity even within the block and will introduce more errors in your analysis. You have enough seeds and large number of genotypes so I would suggest you to use alpha lattice design.

Alpha-lattice designs are replicated designs that divide the replicate into incomplete blocks that contain a fraction of the total number of entries. Genotypes are distributed among the blocks so that all pairs occur in the same incomplete-block in nearly equal frequency. The design permits removal of incomplete-block effects from the plot residuals and maximizes the use of comparisons between genotypes in the same incomplete-block.

My colleague Pablo Díaz-Rueda is an expert in working with more than 100 genotypes. He published it in those papers:

Article SILVOLIVE, a Germplasm Collection of Wild Subspecies With Hi...

And this one

Article Hydraulic Traits Emerge as Relevant Determinants of Growth P...

Article Evaluation of early vigor traits in wild olive germplasm

Dear Researchers,

Thank you all for valuable suggestions. After reading all the comments, I came to know that Alpha Lattice would be appropriate for my experiments.

Thanks alot.

The augmented design is most appropriate with this number of genotypes. Every 5 or 10 genotypes sown in rows (one or two) should be intercalated with the control sown in one or two rows too. It is the check that is repeated.

Thanks

Augmented design should be used for this. In this design checks are replicated and it should not less than 4

Alpha lattice design will be appropriate as you have a large set of varieties which you want to grow at the same time.