Research Data Life Cycle

Before collecting data, it is best to plan ahead and ask yourself: What types and formats of data will be collected? Is there any copyright issue involved? What are the best approaches to store and back up data?
You may go to the Research Data Management library guide for more information.

Data can be collected…

• • Through observation – generally be collected once and is unique

  • By experimenting – through experiments; in general can be repeated

  • By simulation – test models; usually can be reproduced

  • By researching sources – deriving from literature, manuscripts, publications, etc.

  • By data processing – combining, reprocessing, (re)grouping, etc. of data created before

  • By using existing data

This step involves data inputting (if the raw data is not collected in a digital format), data conversion (from one system to another system, or from one format to another format), and data cleaning.

Data cleaning requires tedious and time-consuming manual work, but its importance should not be underestimated. Proper data cleaning can prevent researchers from coming back to this step at a later stage of the research and avoid drawing false conclusions.

The following data cleaning tips can serve as a starting point:

• • Clear field labeling – make sure you can understand the labels even after one year of time

  • Remove unwanted observations – including duplicate or irrelevant observations

  • Filter unwanted outliers – only for the suspicious measurements that are unlikely to be accurate

  • Handle missing data – by dropping observations with missing values or inputting missing values based on other observations

  • Fix structural errors – including typos, inconsistent capitalization, and inconsistent name formats

  • Controlled vocabularies may help – e.g., develop a small dictionary to remind yourself to use "United States" (instead of "USA" or "America") or "computer" (instead of "computers" or "PC") throughout the document

  • Beware of strange characters – especially when you directly copy and paste web contents into an Excel; an invisible strange character is usually added at the end of a sentence

Some of these points are mentioned in EliteDataScience. Go there for a more comprehensive explanation.

This is the most challenging but also most exciting part of the cycle. It can involve quantitative analysis, qualitative analysis, machine learning, etc.

This guide does not intend to cover basic statistics that can be found on the Internet easily. (If you have no idea which internet sites to use, you may start with Statistics How To.) We hope to introduce commonly-used software tools instead.

This step involves short-term measures such as proper file version control during a research project and long-term data archiving measures to migrate data to the best format and store it in the most suitable medium for your or your company's future use. You may learn more about this through TechTarget.

Data storage is more on internal use of data, but data sharing refers to open data that can be accessed and re-used by the public for free. Open data is not only a trend but also an obligation that researchers are recommended to meet for the benefits of academia and the society. Some major publishers also request authors to share their data, e.g., Nature and Science.

Data can be shared in its original form (after removing privacy and sensitive information) through publicly accessible data repositories. Researchers can also choose to share their data through data visualizations or developing interactive web applications.

There are many free and subscribed data resources available for researchers to re-use. We have prepared another library guide for data resources, please visit the guide on Sources for Data-Mining.