|Moderator:Jouni (see all)|
|This page is a stub. You may improve it into a full page.|
This page is a brief guidance for developing variables. Links lead to more detailed descriptions of relevant issues. This page assumes that the topic is already known. For a help about how to clarify an assessment and the topics for the variables needed, see Developing assessments. In addition, this page gives guidance on how to get started, not on how very mature variables can be further developed.
What is a practical way of developing variable pages so that a new participant gets useful work done without highly specified skills?
Based on long experience.
- Think about your topic and clarify it as an unambiguous research question.
- Look for other similar variables.
- Is your variable actually a single study or otherwise a piece of original data with a permanent nature? If yes, create a study page, otherwise create a variable page.
- Write down your question.
- Go for the first quick round for data about your variable. Write down any links or short descriptions of data that may help you answer the question. Focus on easy sources of information.
- Based on the information collected, make the first estimate of your answer to the question.
- Think about the smallest and largest plausible answer, and rationale why the answer might be so small/large. Plausible means that you are 95 - 99 % sure that the true answer is inside the range.
- Write down your range of plausible answers (to the section result) and the rationale (to the section definition).
- Upload your answer to Opasnet Base. You can use the index Parameter to describe the min and max of your range, or you may use a distribution code that is understandable to R. For help about uploading data, see Opasnet base connection.ANA.
- Think about things that might be causally related, i.e. things that may change the result of your variable if they change. List these things under the subheading Dependencies. Create new variables for these things as necessary.
- Describe the identified dependencies in a quantitative manner if possible. You can use conditional probabilities or functions. However, start with simple descriptions such as linear functions or statements like "if the result of the parent variable is high (i.e., above x), the result of the variable will be between y1 and z1; if it is low, the result will be between y2 and z2." You can also estimate rank correlations between the parent and the variable.
- If you have successfully done the previous steps, you can remove the stub=Yes parameter from the variable template and vote about the quality and usefulness of the page.