This is in the clinical psychology field, but any research guidelines I could reference would be great!
I believe there is a move towards this way of working: where someone who has no vested interest in the results, no bias, as they are not part of the intervention team, is chosen to perform the data analysis.
However, I am struggling to find any official recommendations backing this approach. Does anyone know of any? Many thanks!
It's important that the analyst has an analysis plan. This sort of independent analysis is common in drug trials. You should have a look there.
The important things are that the data are blinded so that the analyst does not know the identities of the groups being compared (in trials the treatments are just given letters). It's also important that the analyst should stick to the analysis plan. There are times when this is impossible (not enough data, or the data clearly violate the assumptions of the specified model) in which case they can use an alternative.
In drug trials it is also common for a second analyst to inspect the analysis to see if the report is done according to the analysis plan, and all relevant checks specified have been made. At that point the report is unblinded.
Hi,
In a clinical trial, triple blinding can also be done and is included in the protocol. The statistician doing the analysis is blind and is given coded variables along with the statistical plan. Later, the clinical team interprets the analysis.
The statistician usually asks to be put as a co-author. In psychology this has been no problem.
- Badges
- Science topic