Skip to main content

What is Data Quality, and How to Enhance it in Research

 



We often talk about “data quality” or “data integrity” when we are discussing the collection or analysis of one type of data or another. Yet, the definition of these terms might be unclear, or they may vary across different contexts. In any event, the terms are somewhat abstract -- which can make it difficult, in practice, to improve. That is, we need to know what we are describing with those terms, before we can improve them.

Over the last two years, we have been developing a course on Total Data Quality, soon to be available on Coursera. We start from an error classification scheme adopted by survey methodology many years ago. Known as the “Total Survey Error” perspective, it focuses on the classification of errors into measurement and representation dimensions. One goal of our course is to expand this classification scheme from survey data to other types of data.

The figure shows the classification scheme as we have modified it to include both survey data and organic forms of data, also known as big data or found data. We find that all forms of data are subject to these same sorts of errors in varying degrees.

We won’t define all the classes in this post – just two examples.

Data Origin

First, on the measurement side, we look at “Data Origin” as how were the individual values / data points for a given variable (or field) recorded, captured, labeled, gathered, computed, or represented? This could be the process of answering a question, filling a field in an administrative record, or labeling an image in a machine learning context. In the case of labeling images, this could be a human being incorrectly labeling an image. For example, a human being might not note the difference between a cat or a kitten. In some contexts, that difference could be important.

Missing Data

On the representation side, “Missing Data” is a common problem that impacts many types of data. For example, administrative records can be missing key variables or even entire records. Similar things can happen with surveys. These missing data can impact inferences or predictions if the missing values differ from the observed values in important ways.

Using this classification scheme as a way to think about errors can help guide researchers as they consider quality issues. Further, being aware of these issues may also open the door to enhancing the quality along these dimensions! If you’d like to learn more, our new open online courses series focuses on identifying, measuring, and maximizing quality along all of these dimensions.

This blog post is also published here.




Comments

Popular posts from this blog

Slowly Declining Response Rates are the Worst!

I have seen this issue on several different projects. So I'm not calling out anyone in particular. I keep running into this issue. Repeated cross-sectional surveys are the most glaring example, but I think it happens other places as well. The issue is that with a slow decline, it's difficult to diagnose the source of the problem. If everything is just a little bit more difficult (i.e. if contacting persons, convincing people to list a household, finding the selected person, convincing them to do the survey, and so on), then it's difficult to identify solutions. One issue that this sometimes creates is that we keep adding a little more effort each time to try to counteract the decline. A few additional more calls. A slightly longer field period. We don't then search for qualitatively different solutions. That's not to say that we shouldn't make the small changes. Rather, that they might need to be combined with longer term planning for larger changes. That

An Experimental Adaptive Contact Strategy

I'm running an experiment on contact methods in a telephone survey. I'm going to present the results of the experiment at the FCSM conference in November. Here's the basic idea. Multi-level models are fit daily with the household being a grouping factor. The models provide household-specific estimates of the probability of contact for each of four call windows. The predictor variables in this model are the geographic context variables available for an RDD sample. Let $\mathbf{X_{ij}}$ denote a $k_j \times 1$ vector of demographic variables for the $i^{th}$ person and $j^{th}$ call. The data records are calls. There may be zero, one, or multiple calls to household in each window. The outcome variable is an indicator for whether contact was achieved on the call. This contact indicator is denoted $R_{ijl}$ for the $i^{th}$ person on the $j^{th}$ call to the $l^{th}$ window. Then for each of the four call windows denoted $l$, a separate model is fit where each household is as