Curated Datasets: Great for Data Science Portfolio Projects!

Pros:

• Well-documented
• Datasets free to download online
• Some extremely useful questions for analysis
• Goes many years back so trending possible
• Core dataset includes representative sample of US population by states and nationally, including weights for weighted analyses
• Large (over 400,000 records)

Cons:

• Some outdated questions
• Lack of contemporary questions (e.g., e-cigarette use)
• Cross-sectional, so cannot be used for longitudinal studies
• Recent severe reduction in core questions asked (as of 2020)
• Lacks clinical data or data that can be measured in-person (see NHANES)

Pros:

• Very easy to browse and download data using the dashboards
• Well-documented
• Up-to-date
• Even if you do not have a background in pharmaceuticals, it is easy to look up information about drugs, the US, and other topics to be able to come up with a question to answer using some of these data.

Cons:

• The data served up are not necessarily the data you want. There are calculated variables that apparently have to do with seeing if a policy is met that are not useful generally for analysis (e.g., counts of people aged 65 and over receiving a drug vs. everyone).
• Even though it is easy to acquire the domain knowledge necessary to do an analysis, it is quite consuming. Drugs are complicated, Medicare and Medicaid are complicated, and US healthcare finance is also complicated.

Pros:

• Well-documented
• This is a very high-quality dataset. The measurements seem very valid and reliable.
• Very long-term follow-up, so survival analysis models can be used
• For the entire trial the sample size is pretty big, and the data appear to be pretty clean.
• Datasets are available at no cost but you must undergo authorization

Cons:

• Very old-fashioned dataset and structure. Non-relational, so very confusion.
• Documentation page has all necessary information, but is clumsy, slow, and labyrinthian.
• Need a lot of domain knowledge to understand the dataset. Measurements such as psychological instruments, labs, and genetics require advanced knowledge to analyze.

Pros:

• Date are current and up-to-date
• Easy to download in ASCII format free over the internet
• Well-documented – zip file contains explanation of each dataset
• Although somewhat confusing, it is not that challenging to understand the data from the point-of-view of domain knowledge
• Very nice dashboard available to help you understand the data before you download it.

Cons:

• Data are biased, in that many adverse events go unreported
• Data need to be aggregated to be meaningful. For example, if you want to study “Lisinopril”, or “hypertension medication”, or certain classifications of adverse events, you will need to look at all the different formulations and develop classifications.
• Multiple tables, so it will take some reading of documentation to figure out what you want to do with this dataset
• No load code posted

Watch our livestream recording for a demonstration of how to query the public FAERS dashboard.

Pros:

• Well-documented
• Goes many years back so trending possible
• Includes rare cost-related data so can be used for healthcare cost forecasting
• Includes many calculated variables based on economics that are useful in forecasting models

Cons:

• Datasets are for sale, and although prices are not that high, it takes a lot of research on their documentation web site to figure out what data you want to purchase
• Due to lack of access in the US healthcare system and other features of US healthcare financing and delivery, the dataset is unbalanced and biased, and it is hard to use for health-related analyses. In fact, drawing inferences may actually perpetuate bias in the US healthcare system.
• Many variables in the dataset are very hard to understand. They require knowledge of both the healthcare setting as well as healthcare economics.
• Data prepped for SAS, but not open source applications

Pros:

• Extremely well-documented and well-used
• Goes many years back so trending possible
• Analyses extremely helpful to the functioning of the US military
• Servicemembers and veterans who want to do epidemiologic analyses: These are GREAT datasets for you, because you will already understand them, and will have less trouble getting access to them for projects

Cons:

• Requires a lot of domain knowledge. If you have not worked for the military or been in the military, these will be very confusing datasets for you.
• Difficult to get permission to access unless part of a research team that has approval. A good way to address this is to become part of a research team that gets a grant from the military or other organization to use these datasets to study the military.

Pros:

• Datasets free to download online
• Some extremely useful questions for analysis
• Goes many years back so trending possible
• Population-based sampling with weights
• Includes rare in-person examination data, such as oral health data, and anthropometrics.
• Also includes some laboratory measurements, and other hard-to-measure health data.

Cons:

• Some outdated questions
• Lack of contemporary questions (e.g., e-cigarette use)
• Cross-sectional, so cannot be used for longitudinal studies
• Dataset is small (<10,000 per year). For a larger health survey with similar measurements, see BRFSS.
• Extremely fragmented dataset, in that each exam or questionnaire is stored in its own dataset, so hard to put together a dataset with all the covariates needed about every experimental unit. For this reason, many researchers aggregate multiple years of NHANES data to answer research questions.
• Data prepped for SAS, but not open source applications

Watch our livestream recording for an introduction to the NHANES documentation and guidance for downloading and using the data.

Pros:

• Very large datasets each year (screening and topic are over 50k rows).
• National stratified cross-sectional sampling in the United States (including weights).
• Data available online back to 2016
• Trustable, updated and well-documented.

Cons:

• May be hard to estimate across entire child population because there are three topic questionnaires: T1 for age 0 to 5, T2 for age 6 to 11, and T3 for age 12 to 17
• Screening dataset doesn’t have a lot of different columns
• Requires quite a bit of background research before you can devise an analytic plan
• Not used as often in analyses so not a lot of precedent set

Pros:

• Voluminous, accurate data from a trusted provider.
• It is easy to choose data from companies a la carte – so you can focus in a warranty domain (e.g., automotive, energy, manufacturing).

Cons:

• You need to have a lot of domain knowledge to understand this dataset and figure out what to do with it. If you want to know more, please contact Joe Chantiny on LinkedIn and he can give you more information.
• The spreadsheets are NOT tabular – each spreadsheet is essentially a report. To do anything with the data, you would have to process it into data tables.
• Data are not free, but the cost is low, and you can pay by PayPal. However, you need to shop carefully on the site to decide exactly what you are buying. They have 394 datasets, and as of this writing, to purchase all of them is less than $2,000, which is actually a great deal in my opinion.

Pros:

• If you are a casino customer, it is very easy to understand the data.
• Interesting to use in time trending, and could be useful if paired with casino data from nearby states to review competitive pressure as the different casinos in Massachusetts opened.
• Great examples of continuous variables and monetary data for practice.
• Opportunity to connect to other datasets about Massachusetts for richer information
• Portfolio projects can be very interesting, flashy, and easy to understand

Cons:

• Electronic datasets need to be developed – they are in PDF format. It is not hard to do data entry, but an enterprising programmer could scrape the data easily as well.
• Small datasets, so very little data. Casinos have not been open in Massachusetts very long.
• Only a few columns, so you need to be creative to develop research questions and analyses to answer them.

Pros:

• They have a new rebuilt portal (goodbye American FactFinder)!
• These data are very helpful when you are writing about a specific organization or location in the US. If I am analyzing data that came from students in a particular school, then I can use census data to characterize the region around that school.
• If you understand how counties work in the US, and you are analyzing other data that has county variables (e.g., data from hospitals located in certain counties), then “hooking on” county-level estimates is a great way to add value to your data from the census.

Cons:

• You really have to understand the US in terms of geographic regions in order to utilize the data. You need to understand states, counties, MSAs, census districts, and other regional definitions.
• Some data are suppressed for privacy reasons.
• It’s great that there are so many datasets and so much data available, but many are hard to understand. Beyond the simple analysis, it will take you some time to get to know what exactly is available, and what you can use for your purposes.

Pros:

• Very easy to start using the platform. Very intuitive user interface.
• Would not be difficult to design an ecological/correlational study with the data based on some sort of hypothesis about voter turnout and country-level policy.
• Trustable, updated and well-documented.

Cons:

• Data on many metrics are missing from countries that are probably the most important to track for democracy purposes.
• Political systems are different in every country, so it actually is hard to compare.
• Only reflects national elections – but even those can be complicated (e.g., president vs. parliamentary election).
• Not obvious how to download datasets (for putting in statistical software).

Pros:

• Extremely thorough and comprehensive data and documentation.

• Data structured to be used in a statistical program (like R or SAS).

• Dataset subject of peer-reviewed article and extensively vetted.

Cons:

• Requires extensive study of different countries and the data available to formulate a research aim.

• Must apply to obtain access to dataset.