Table 3 Replication experiments stages

1. Determine maximum number of replicates that can be performed in a single day

Four replicates are the minimum to calculate within-day imprecision. This number also accounts for potential outlying observations

2. Define total number of tests based on desired number of replicates and feasibility

Perform 20–30 replicates unless (Westgard, 2020)

• Published data or

• Preliminary development results indicate that 30 replicates are too few to evaluate measurement outcome

Ensure sample size is sufficient to capture any downstream measurement outcomes, e.g., fecundity

Implement a balanced design where applicable. Here, the number of control replicates should equal the number of test replicates. If a balanced design is not practical, a larger proportion of the sample size should be allocated to treatment arms

Refer to ‘Sample size’ section above for guidelines to determine sample sizes

Replication experiments are performed for the entire defined reportable range (minimum 4 points with 20–30 replicates tested for each), where appropriate

3. Perform experiments over the smallest possible number of days

A minimum of four replicates should be tested per day for a minimum of 5 days in replication studies (holding all the testing conditions/factors constant) to allow estimation of the different levels of precision [21, 64]

Where possible a balanced (equal number of replicates) design should be employed

4. Results from single days of testing are used to estimate within-run or day variation for each measurement outcome

If multiple test days were considered, pooled within-run or day variance estimates are be calculated

5. Results from multiple days of testing are used to estimate between-run or day variation for each measurement outcome

Calculate the between-days variance for the multiple testing days

6. Whole dataset provides estimate of variance for sample size calculations for total replication study

Used to estimate the total error for an entire experiment