Difference between revisions of "2.1.5 Pre-specification"

From EQIPD
Jump to: navigation, search
(Created page with "== ​​​A. Background & Definitions == Pre-specification is a protective measure against so called "rationalization" („making excuses“), a defense mechanism in which...")
 
Line 55: Line 55:
  
  
Precommitment (Wikipedia article) - link
+
Precommitment (Wikipedia article) [https://en.wikipedia.org/wiki/Precommitment]
  
  

Revision as of 13:48, 2 September 2020

​​​A. Background & Definitions

Pre-specification is a protective measure against so called "rationalization" („making excuses“), a defense mechanism in which controversial or questionable behaviors or feelings are justified and explained in a seemingly rational or logical manner to avoid the true explanation, and are made consciously tolerable - or even admirable and superior - by plausible means (e.g. "I didn't get the job that I applied for, but I really didn't want it in the first place").

An example of a situation when rationalization mechanisms can get engaged:

  • You are running an experiment and there is a sudden noise, vibration or some other factor
  • You get angry but cannot do anything other than recording in your lab journal that such event occurred at this particular time point
  • When analyzing the data, you notice that one of the subjects / data points behaves strange and it is exactly when that disturbing interference occurred
  • If including or excluding this subject or data point makes the results appear differently, what do you do?

Pre-specification aims to protect scientists from giving a wrong answer to the question above and, thereby, supports generation of unbiased results.​

Pre-specification is mechanistically similar to precommitment or self-binding when we make the decision before being in the tempting situation (e.g. take a limited amount of money with you to curtail spending; have only healthy foods at home to avoid the temptation to go astray).

Thus, pre-specification is about a strategy that we may use to restrict the number of choices available to us at a future time.


B. Guidance & Expectations​

EQIPD advises to pre-specify key study and analysis details irrespective of the purpose of research.

For knowledge-claiming research (Purpose of research), pre-specification is mandatory and should cover all key study and analysis details as well as the key decisions, e.g.:

  • study hypothesis
  • data analysis plan (including statistical methods to be applied)
  • inclusion and exclusion criteria (including criteria for "outliers")
  • primary outcome measure
  • if applicable, interpretation of the study results dependent on performance of the control groups or treatments (e.g. is the study declared "failed study" if a positive control fails)
  • if applicable, decisions to be made if the primary outcome is met or not met

It is recommended that pre-specification is documented by the same scientist who prepares and documents the Study (experimental) plan.

Pre-specification requires no special training and can be completed as a free-text summarizing the study and analysis details that are to be pre-specified.

Information about pre-specified study design and analysis details should be saved in a manner that:

  • identifies the author and time of creation
  • is protected against deletion and unauthorized modification
  • is retrievable and readable

The most optimal way of performing pre-specification is by storing pre-specified information in a laboratory notebook (e.g. electronic) as part of a Study (experimental) plan​.

Pre-specification may also be conducted by means of Preregistration of the study plans.

Any amendments to pre-specified study design and analysis details should be properly documented (dated, signed, etc.).

PLEASE DO NOT FORGET

To check whether the means chosen for pre-specification ensure safe and secure storage of pre-specified information ​To consider adding this subject to an awarenss training program for new employees or refresher training


C. Resources

Precommitment (Wikipedia article) [1]



back to Toolbox

Next item: 2.1.6 Sample size and power analysis