How to Use EDA to Investigate the Relationship Between Public Housing Availability and Homelessness

Exploratory Data Analysis (EDA) is a critical first step in understanding the underlying patterns, relationships, and trends in a dataset. When investigating the relationship between public housing availability and homelessness, EDA helps to uncover key insights and potential causal connections. By analyzing the data visually and statistically, you can identify correlations, distributions, and outliers that may reveal patterns in how public housing availability affects homelessness. Here’s a structured approach to use EDA for investigating this relationship:

1. Define the Problem

Before diving into the data, clearly define the research question: how does public housing availability affect homelessness rates? This requires gathering data that measures both the availability of public housing and homelessness statistics.

Key Questions to Consider:

• How is public housing availability measured? (e.g., number of available units, housing affordability, waiting lists)

• What data is available for homelessness? (e.g., number of homeless individuals, homelessness rates, demographics)

2. Gather and Clean the Data

The next step is to gather relevant data sources. The two key components are:

• Public Housing Data: This might include data on available units, waiting lists, budget allocations, and housing policies.

• Homelessness Data: This could include the number of homeless individuals, shelter capacity, emergency housing programs, and demographic information.

Once you have the data, the next step is cleaning. This involves:

• Removing missing or incomplete records.

• Converting data types (e.g., ensuring that dates are in proper format).

• Handling outliers or inconsistencies in the data.

• Aggregating or disaggregating data where necessary (e.g., monthly, annual, or by region).

3. Univariate Analysis

Start with analyzing each variable in isolation to get a sense of its distribution and key statistics.

• For Public Housing:

  • Visualize the distribution of available units over time (e.g., a line plot or bar chart).

  • Summarize key statistics like mean, median, standard deviation, and range.

  • Identify trends or seasonality in housing availability.

• For Homelessness:

  • Plot the distribution of homelessness across different regions or time periods.

  • Check the statistics (mean, median, etc.) to see if the number of homeless individuals has changed significantly over time.

These initial analyses will help you spot any unusual trends or patterns in the data.

4. Bivariate Analysis

Now, the real work begins: examining the relationship between public housing availability and homelessness. Several approaches can be used here:

Correlation Analysis:

• Correlation Coefficient: Start by calculating the Pearson correlation coefficient between the number of available public housing units and the number of homeless individuals. A positive correlation suggests that more public housing is associated with fewer homeless individuals, while a negative correlation indicates the opposite.

• Scatter Plots: Visualize this relationship using scatter plots. You can plot the number of available public housing units on the x-axis and the number of homeless individuals on the y-axis. This can give you an immediate sense of whether there is any visible relationship.

Grouped Analysis:

• Time-Based Analysis: If you have time-series data for both homelessness and public housing availability, you can use line graphs to visualize how changes in housing availability correlate with fluctuations in homelessness over time.

• Regional Comparison: You can also compare different geographic regions (cities, states, or countries) to see how public housing availability and homelessness rates vary. Use box plots or bar charts for this type of analysis.

Categorical Analysis:

• Demographic Breakdown: It can be useful to break down homelessness data by demographics (age, gender, race, family status) and compare these breakdowns with public housing availability. Cross-tabulations and stacked bar charts are useful for visualizing this.

5. Check for Confounding Variables

Often, other factors can influence both public housing availability and homelessness, such as income levels, unemployment rates, or social services. To ensure that the relationship you observe between public housing and homelessness is not confounded by these other variables, you can:

• Conduct multivariate analysis (e.g., regression analysis) to control for other factors.

• Use heatmaps to visualize correlations between multiple variables (e.g., income, unemployment, housing availability, and homelessness).

6. Trend Analysis

You can use time-series analysis if your data spans over a significant period. Look for trends in public housing availability and homelessness over time. For example:

• If public housing availability has increased, has there been a corresponding decrease in homelessness rates?

• Use moving averages or seasonal decomposition to understand trends and seasonal variations in both datasets.

7. Outlier Detection

Outliers are unusual data points that deviate significantly from other observations. These might represent unique situations that require further investigation. For instance:

• An unexpected drop in public housing availability in a specific year could correlate with a sudden increase in homelessness.

• Detect and analyze outliers to see if they are due to data errors or represent real-world anomalies.

8. Visualization

Visualization is an essential part of EDA because it helps convey complex relationships clearly. Some useful plots include:

• Heatmaps for correlation matrices.

• Pair plots to explore relationships between multiple variables (e.g., homelessness rate, housing availability, unemployment, etc.).

• Box plots to show the distribution and variation of public housing availability and homelessness across different categories or regions.

9. Hypothesis Testing

EDA often leads to hypotheses about the data that you can test statistically. For example:

• Does a city with more public housing units have a significantly lower rate of homelessness?

• Are regions with higher unemployment rates also seeing a higher rate of homelessness, even after controlling for housing availability?

You can perform statistical tests such as:

• T-tests or ANOVA to compare differences in homelessness rates across different levels of public housing availability.

• Chi-square tests for categorical data to explore relationships between homelessness and different demographic groups.

10. Conclusion and Further Investigation

After completing your EDA, you should have a clearer picture of how public housing availability impacts homelessness. If your analysis shows a significant relationship, you can move forward with more detailed modeling (such as regression analysis) to quantify the impact of public housing on homelessness rates.

Additionally, your findings may lead to new questions or hypotheses for further investigation. For example:

• What specific types of public housing (e.g., affordable housing, subsidized housing) are most effective in reducing homelessness?

• Are there other factors (e.g., mental health services, substance abuse programs) that interact with public housing availability to affect homelessness?

By using EDA effectively, you can gain actionable insights that inform policies and interventions aimed at reducing homelessness through increased public housing availability.