Motivation

Background

Data

What Does This Tool Visualize

The Source of The Data and Biases and Ethical Issues Embedded In The Data

This real estate dataset was collected by LINKMV, a Cape and Islands based real estate multi-listing service (MLS) provider. The purpose of a multi-listing service is to work directly with real estate brokerages to aggregate real estate sales into a single unified data service. This dataset, therefore, contains information about real estate sales in the Cape and Islands.

One important point about this dataset is that it is relatively unbiased because LINKMV is a neutral third-party to the real estate market. However, it's worth noting that bias may be introduced in the form of the "Description" attribute, which is provided by the brokerages themselves. These brokerages have a profit incentive to sell more properties and may therefore provide descriptions that are overly positive or even misleading.

In terms of ethical considerations, this dataset provides granular information about people's homes, including the sale price, location, and features of the property. While this information is publicly available, a certain degree of inspection must be carried out by the brokerages, the MLS provider, and the data team to ensure that no sensitive information makes its way into the dataset. Sensitive information could include things likethe personal contact information of the property owners, or details about the property that could be used to identify the owner. It's important to ensure that this information is protected and not made public in the dataset.

Another important consideration when using a real estate dataset is the potential for spatial autocorrelation, which occurs when nearby observations are more similar to each other than they are to distant observations. In the context of real estate data, this means that the sale prices of nearby properties may be more similar to each other than they are to sale prices of properties located further away. This can be an issue when using statistical models that assume independent observations, as it can lead to biased estimates of coefficients and standard errors. To account for spatial autocorrelation, techniques such as spatial regression or spatial filtering may be used.

Another consideration when working with real estate data is the potential for missing or incomplete data. Incomplete data can arise from various sources, such as data entry errors, data suppression due to privacy concerns, or missing information on certain attributes of a property. Missing data can have an impact on the accuracy of statistical models, as it can lead to biased estimates of coefficients and standard errors. Therefore, it's important to carefully handle missing data by imputing missing values or using statistical techniques that can handle missing data.

Finally, when using a real estate dataset, it's important to consider the potential impact on communities and individuals. Real estate data can reveal patterns of segregation, gentrification, and discrimination in housing markets, which can have a profound impact on people's lives. Therefore, it's important to use real estate data in an ethical and responsible manner, taking into account the potential social and economic implications of the findings. This may involve working with community organizations or advocacy groups to ensure that the findings are used in a way that promotes fairness and justice in housing markets.

Data Quality Issues Found

Data quality issues discoverd in the data include difficult types of data such as photos and urls, as well as missing values and inconsistent values throughout the dataset. To clean the data, we decided to remove the photos columns and other columns that had an excess of missing data. We also decided to remove columns that included links or urls. We combined the exterior features columns to create a singular column with all the exterior features, formatted with commas separating the values. We also re-formatted missing values from NaN's to empty strings, so as to have ease in concatenation, without resulting in a loss of information. Missing values were not filled in with averages or other such methods, as the data missing involved attributes that would be displayed as information about a listing, and weren't major attributes like price or location. Missing values such as “last renovated date” cannot be found through mathematical derivation, and therefore needed to be treated as information that needs to be left out of the additional information portion of the listing. In terms of derived columns, creating a column describing price category allows us to organize the listings into 6 different groups based on the listing's price ranges. This provides the ability to create the heat-map of the different areas of Martha's Vineyard with certain price ranges to complete part of our specified domain tasks. Other simple reformatting of columns was performed for consistency, and reordering of columns was needed so that the information being presented felt organized when looked at in its tabulated format.

Acknowledgements