Table 5. Example of strategies to implement masking during data analysis.

We have classified the masking options as either low, moderate, or high based on the ability to minimise risk of bias. Shown in bold is the recommended high-quality strategy that is readily implementable across different experiment types.

Strategy	Additional detail	Strengths and weaknesses identified by the authors or the literature	Quality of strategy
Independent analyst	The analyst could vary in their independence. Examples vary from a student in the same lab to an independent statistician.	Weakness: • While not emotionally invested in the outcome, the analyst is still fully aware of the significance of 0.05 and the needs of the researcher. • Power dynamics could be at play particularly when the analyst is junior and part of the same department, which will decrease the effectiveness of the strategy to reduce bias	Low-moderate
Randomly code the experimental groups	For example: In a study with four experimental groups (e.g., control, low, medium, high dose), the groups could be recorded randomly to group W, group X, group Y, and group Z.	Weakness: Can still see the experimental group differences. Strength: Simple to implement.	Moderate
Independent analyst and randomly code the experimental groups		Strength: Effective and readily implementable across different experiment types.	High
Adding noise	Add noise (from an appropriate statistical distribution) to the outcome measure to hide the true relationship between intervention and outcome measure.	Weakness: • Precise amount being added is important to be effective and not alter the properties of the outcome measure. • Requires statistical and computational sophistication.	High
Decoy data analysis	Analysts works with multiple data sets (e.g., 6), one of which is the original.	Weakness: • Many of the masking strategies rely on knowledge of what matters in the data, aspects of the data might be unexpected, and, therefore, implementation might eliminate existing effects, induce effects, or change directions of effects [45]. • This strategy does increase the analysis time. • Requires computational sophistication	High
Shuffle the key variables	For example, in a correlation analysis, a scientist shuffled the outcome measure but kept the relationship between independent variables of interest [46].	Weakness: Requires computational sophistication.	High
Adding cell bias to equalise the means	Hide experimental group differences by adding the same hidden number to all observations within an experimental group, which leaves the distribution intact but obscures the differences [42].	Weakness: Requires computational sophistication.	High