bbdaniels
commented
4 years ago Very well written overall and a great improvement. Almost no more content is needed but a little bit of structural cleanup is. The big need is to make the flow and logic a little more accessible, as in to clear up why exactly the workflow is:
- Tidy
- Clean, including HFCs
- De-ID
The info is there but the following points need emphasis:
- "Data points" are different from "data". When we tidy, we change the data from the structure it was collected into the structure that it represents, without changing any information in the data points.
- The desired data structure is known from the data collection instrument and study design (link back to new Ch 3-4), so tidying can be set up before any data is collected and then run as the first step of data intake.
- Quality control is done as an ongoing process during data intake. It makes sure that all real information is represented accurately in the data points.
- Cleaning is done after data intake. It makes sure all information and meta-information is represented and linked to the final data table and the data collection instrument.
- De-identification is done once all the information and meta-information are saved. This process removes information and meta-information that directly identify individuals in the data or involved in the data collection.
So perhaps the "preparing data for analysis" can become an introduction section that summarizes all this right at the beginning? The "correcting data points" subsection there should be part of "data cleaning" in my opinion. Small notes attached in annotations.
luizaandrade
kbjarkefur
Ok, here it is. I'm sure there's still a lot of adjustments to be made, but a fresh set of eyes would help. I went back and forth with the order of sections, so that may be something to pay particular attention to