CONTEXT: It is becoming increasingly important to pre-specify analysis to avoid any p-hacking or data mining. PAPs are more common in the context of randomized control trials (RCTs), but are less common in observational studies. A nascent literature is grappling with the concept of pre-analysis plans for observational studies . Burlig (2018) [and here] also provides a great framework for doing this.
Given that we are still compiling the data, and have thus not technically "seen it", i.e. we haven't analysed the data or seen the distributions of the variables in the datasets.
We can then develop a strong theory behind the academic manuscript and article, publish the PAP on OSF, and then work on data analysis.
We can include "pre-specified analysis" and "exploratory analysis" in the paper, being clear to the readers what analysis was conducted on an ad hoc basis.
tjbrailey
commented
2 years ago
CONTEXT: It is becoming increasingly important to pre-specify analysis to avoid any p-hacking or data mining. PAPs are more common in the context of randomized control trials (RCTs), but are less common in observational studies. A nascent literature is grappling with the concept of pre-analysis plans for observational studies . Burlig (2018) [and here] also provides a great framework for doing this.
Given that we are still compiling the data, and have thus not technically "seen it", i.e. we haven't analysed the data or seen the distributions of the variables in the datasets.
We can then develop a strong theory behind the academic manuscript and article, publish the PAP on OSF, and then work on data analysis.
We can include "pre-specified analysis" and "exploratory analysis" in the paper, being clear to the readers what analysis was conducted on an ad hoc basis.