🏥 Analyzing Hospital Data with SQL

A few years ago, my spouse and I took our kids to meet other family at the Georgia Aquarium. We splurged for a hotel so that we could stay longer at the aquarium.As he was in the bathroom, our two year old started crying. Turns out, he had a rectal prolapse. He describes it as, "My body came out of my body."When you are away from home and a child has a medical emergency, you just expect and hope there is quality health care available.This project write up has not yet been completed.
Last update: June 29, 2026

Why I Chose This Project

Health care is an important industry. You never know when an emergency will make someone need health care, like the story of my youngest son above. You also don't know when you will need a hospital for ongoing care, like my mom did when she was diagnosed with cancer when she was younger than I am now.The health-care system in western North Carolina, USA needs help. The only level 1 trauma center in Asheville is currently under enhanced monitoring by the U.S. Centers for Medicare & Medicaid Services (CMS). CMS has issued at least four immediate jeopardy warnings to Mission Health in the past five years.I chose this project using health-care data to demonstrate that I have the skills and understanding to both add value to a health-care company and add to the well-being of a community.

.

What You'll Learn

You will learn
⚫
⚫
⚫

Key Takeaways

⚫
⚫
⚫

Dataset Details

• Dataset Link - UC Irvine: Diabetes 130-US Hospitals for Years 1999-2008

The dataset contains 10 years of information patients with diabetes from 130 hospitals. The dataset has two tables.One tables has over 100,000 rows with each row being an inpatient experience for a person with diabetes. It has 47 columns with data on the diagnoses, labs, medicines, length of hospital stay, and other health information.The second table has over 70,000 rows of demographic data in 5 columns.There are values that are missing.The dataset comes from the UC Irvine Machine Learning Repository. For more details on how the data was obtain, see Impact of HbA1c Measurement on Hospital Readmission Rates: Analysis of 70,000 Clinical Database Patient Records

Audience

With any data analysis project, it is important to know who the audience of the final report will be. I had hospital administrators in mind as I worked on these analyses.


Since this is a part of my portfolio, there will also be explanations for potential employers or anyone who would like to know more about some of my analytical thought processes. Please message email [email protected] or message on LinkedIn if you would like to know more about my analytical thought processes.

Analysis

Problem #1⚫What is the distribution of time spent in the hospital?
⚫Do most patients stay less than seven (7) days?

Insights

Histograms are a good way to show a distribution. Normally, I would use Tableau, Excel, or Desmos to create a histogram. What is someone wanted a distribution of a large data fast?SQL can be used to create a histogram. I wrote code to round the length of each hospital stay to the nearest whole day. Next, I coded to count the total number of stays for each day. Finally, I had MySQL output an asterisk, *, for every 100 stays for each day. I clicked on the output at 7 since the stakeholder wanted to know if most stays were fewer than 7 days.This histogram doesn't have anything fancy. Yet it communicates clearly that most stays were fewer than 7 days and it was made rapidly.Look below for the exact SQL code I wrote.

• Click for a glossary

Problem #2A new Hospital Director wants to understand which medical specialties tend to have more procedures per hospital encounter.⚫ They’ve asked for a list of specialties and with an average of more than 2.5 procedures per encounter.

• Click for interactive bar graph in Tableau Public

Insights

The new director wants to know which specialties have more than 2.5 procedures per encounter. Coming right up.I decided to use mean to calculate the average and not median. Median would mostly give whole numbers. Seeking specialties with greater than 2.5 average procedures implies that I needed precision to the tenths place. I kept in mind to look out for outliers using mean.I wrote SQL query and got a list of all of the specialties listed from highest to lowest mean procedures per encounter. The top specialty was proctology with 4 procedures per encounter. There was only one encounter with proctology. While I have sympathy for the patient that had 4 procedures, 53 labs, and 11 medications with a proctologist (ouch!), I suspect the new director is not as interested in that one patient encounter.I added to my HAVING clause so that I filtered for only specialties that had more than 50 encounters. This new query returned 5 specialties. Look below for the exact SQL code I wrote.I ran queries to check the distribution for each of the five specialties. They all had a range of 0 - 6. With no outliers or skewed distributions, mean remains a good choice for the average.I exported a CSV from MySQL and imported it into Tableau. I created a bar graph for the new director. I sorted so that the greatest mean procedures per encounters was on top. I also added a color scale based on the total number of encounters. In the tool tips, I included the total number of encounters and the total number of procedures for each specialty. You can find these by clicking hereThe new director did not tell me why they want this information. Perhaps they want to start with making sure cardiology has the resources they need because they have y far the most procedures. Maybe they are wanting to start with thoracic surgery to work on reducing the average number of procedures per encounter.One thing I learned in this project was to use HAVING instead of WHERE when filtering aggregated data in SQL. I love learning new things!

Problem #3The Chief of Nursing wants to know if the hospital seems to be treating patients of different races differently, specifically with the number of lab procedures done.

• Click for interactive bar graph in Tableau Public

Insights

Great question, Chief of Nursing. I will be right back with those numbers on labs and race.Data on the race of patients and the number of lab procedures done for them were located in different tables. This meant I needed to use JOIN to combine the tables.Both tables have a field called patient_nbr. This is an ID number that helps keep track of the data for each patient and detaches their data from anything that could identify them. Using ID numbers for patients allows this data to be public without violating the HIPAA rights of the patients.By using inner join, I told MySQL to find where the patient numbers matched. In this case, it did not matter which type of join I used. I used inner join because all data in both tables should be tied to a patient. If there was a situation where a patient had an encounter and their demographics was not recorded, then I would not want their data in this query. Even though it didn't matter this time, I chose inner join to make the query simpler to run again in case there are future errors introduced into the table.My query showed that the number of lab procedures each recorded race were similar. The range was about 41 to 44 labs.Since this Chief of Nursing is hypothetical, I was not able to ask them in what form that wanted a report. I asked several nurses what they prefer. Most said they are used to getting charts with numbers only. So I created a table and a bar graph.Of course, just the races recorded have similar numbers of lab procedures, that does not mean that there is no racial discriminator or systemic racism at these hospitals. This dataset does not have any information for a patient perspective. I recommend a patient survey that can show if patients feel they are being treated differently.

Problem #4My boss asked me to explore whether hospital encounters with more lab procedures also tend to have longer hospital stays.

Insights

I can explore that for you, Boss. May I ask a question before I begin? You are right, I just did ask one. May I ask a clarifying question in addition to this question and my previous one?How specific do you want me to be with this exploration? You don't need a lot of details. How about if I see how long the typical stay is for patients with average, below average, and above average number of labs? Cool and when would you like this report? 30 minutes. I shall do my best!I went straight to my computer and opened MySQL. I wrote a quick query to calculate the mean, the standard deviation, the minimum, and the maximum for the number of procedures. I noticed there was a good bit of range beyond average + standard deviation. That led me to create five [5] categories, even though I asked the boss about three [3].Next, I wrote another SQL query to calculate the average length of stay for each procedure frequency category. I did this using the CASE WHEN structure. CASE WHEN is similar to IF conditional statements that can be used in Excel, Python, R, C#, and many coding languages.I added a third column to the query to display the number of encounters that were in each category. This ensures that the days in hospital averages are not greatly impacted by small samples. This also shows the sample is skewed to the right, meaning there are many more encounters with well below average number of procedures and well above average.This table seemed to show a pattern that patient encounters with more procedures tended to have longer stays in the hospital.I wanted to make a scatter plot and run a regression to check for a trend. I have taught scatter plots to middle and high school students. However, it was almost time to give the boss their report. I am sure they won't mind getting it 5 minutes early.I emailed the table to them with the insights seen above. I told them I am available to explain more if they want. I told them I have some time after lunch to investigate this potential trend.The two images below show the SQL code I wrote for this problem.

Problem #5
You just got an email from a co-worker in Research. They want to do a medical test with patients who meet either of these conditions:⚫ Their race is listed as African American.
⚫ Their metformin value is marked as 'Up' in at least one hospital encounter.They need a clean list of patient IDs as fast as possible. What do you do?

Insights

One clean list coming right up. Would you like that as a CSV or another format? Research does amazing work. I am glad to be able to help out.My co-worker asked about information that is on two different tables. This seems like a situation where I need to JOIN the tables. I could do that. My co-worker just needs one clean list of patient numbers. Both tables have a column titled "patient_nbr." I can combine these tables using UNION.I chose UNION instead of UNION ALL because UNION by itself will filter out duplicates. This will give my co-worker in research their clean list.This was my first time using UNION in a query. I wanted to double check my work before sending it to research. I do not want them to do work on incorrect data. So, I wrote a second query using INNER JOIN. I got the same results!Yay! I love learning new commands and confirming accuracy.Note: The image of the table above is a partial list. I suspect nobody reading this projects wants to see 17,375 patient numbers. If you do, feel free to email me.The two images below show the SQL code I wrote for this problem.

Problem #6
The Hospital Administrator wants to review emergency encounters where the hospital stay was shorter than the overall average.

Insights

Dear Hospital Administrator,Thanks for entrusting me with the task of gathering this information for you. What information would you like about about these shorter emergency encounters? Do you want me to do analysis on these encounters? Is there any specific information about these shorter encounters?The administrator replied they wanted all of the information available in raw form so they could brainstorm over it.The health table has a column titled admission_type_id. The data dictionary did not have documented which id number represented emergency admissions. I asked the database administrator. She informed me that 1 is for emergencies.To compile all of the emergency encounters that were shorter than the average hospital stay, I could have calculated the mean hospital stay length and then filtered by less than that number. What if the administrator wanted this information updated in the future?I wanted to write a query that I could run again in the future that would automatically filter on the newest average hospital stay. That presented a challenge as the WHERE clause in SQL cannot filter by an aggregate, such as average. I overcame this challenge by using CTE, Common Table Expression.I was personally curious what the mean length of stay in the hospital was for all encounters. Thankfully, MySQL Workbench will sort a table when the column heading is clicked. I did this and saw that 4 days was the greatest time in hospital in the results table. This means the average must be greater than 4 days, probably 5 days.Note: The image of the table above is a partial list. I suspect nobody reading this projects wants to see 33,684 rows of data. If you do, feel free to email me.The image below show the SQL code I wrote for this problem.

Conclusion and Personal Reflections

Thoughtful words

Recommendation

Nurses are super important. This wasn't in the dataset. Just writing it now because it is true.

Call to Action

If you're interested in data-driven insights or need a data analyst, I'd love to connect and explore how I can help your business grow. Feel free to reach out!Could your organization benefit from my curiosity-driven analyses? Do you like to interact with a lifelong learner who improves his skills with each project?Let’s connect.
Email [email protected]
Connect on LinkedIn

• Home
• Projects
• About Me
• Resume
• LinkedIN

Glossary

• Top of this Project

Sources

• Top of this Project

🏡Residential Real Estate in the United Kingdom

Latha math! Good day!My grandmother was born in Alexandria, Scotland. This is close to the bonny, bonny banks of Loch Lomond. The photo above shows some property for sale in Alexandria.

Why I Chose This Project

I chose this project as a part of my internship with Snow Data Science. I wanted to do this project because many people would like more information when buying or selling a home. Completing this project help me to grow my skills to give insights to help people.I also chose this project because the data set was messy that required some difficult decisions. I wanted to grow my analyst skills by taking on this challenge.Finally, I chose this project because it came with a team. I worked with four others remotely to answer nine challenging and interesting questions.Unfortunately, there was nothing in the dataset about property in the UK where my family have lived.

What You'll Learn

You will learn about residential real estate trends in the United Kingdom.
⚫You will see how the turnover rate for property changes based on the postcode areas and type of home.
⚫You will see how the price of a home increased in Guildford as you move up the property ladder.
⚫You will see how the price changes based on the postcode area and the tenure of the property.

Key Takeaways

⚫ The median turnover time is 7.4 years. The range of median years has a low of 5.7 years in Turnbridge Wells and a high of 9.9 in Ilford.
⚫ Flats and terraced had about the same turnover rate. Higher rungs on the property ladder had about 10% longer turnover than the previous rung.
⚫ Freehold property costs more than leasehold, with the difference in median ranging from 17% more to 220% more, depending on the area.

Dataset Details

• UK Property Prices with Sale History

The data set was published on Kaggle and originally came from Right Move and HM Land Registry. The data set has two tables linked with a primary key.Both tables contained real data residential real estate deals in the United Kingdom from 1998 - 2025.Some of the data was missing. Some of the data was not formatted for analysis. Some of the data presented challenging decisions that I had to make.

Audience

With any data analysis project, it is important to know who the audience of the final report will be. My supervisor wrote:"Everyday home buyers and sellers in the UK, plus curious listeners who want a clear story about what is happening in prices.So not real estate agents, not government, not commercial investors as the main focus."Since this is a part of my portfolio, there will also be explanations for potential employers or anyone who would like to know more about some of my analytical thought processes. Please message email [email protected] or message on LinkedIn if you would like to know more about my analytical thought processes.

Messy Data Decisions

Same day sales
The dataset included properties that were sold on the same day. One property in Aberdeen was sold five (5) times in one day. I did some research and confirmed these were real sales and not a data error.Everyday home buyers and sellers are not typically buying and selling the same house one the same day. Since this is my audience, I did not consider any properties sold in less than ten (10) days.In Tableau Public, the minimum number of days between sales can be adjusted to meet the needs of anyone using this analysis.Sales Price Outliers
I used medians instead of mean when analyzing sales prices because there were some outliers. This was I could keep those sales in the data without the outlier prices making a large impact on the averages.Aberdeenshire
Property sales in Aberdeenshire were the weirdest. About 1,000 sales did not have a tenure listed. Overall sale prices were significantly lower than all other areas. Plus the property sold five times in one day mentioned above had a sale price of £1,000.The dataset was missing a lot of data from Scotland. It did not have any sales in Glasgow. For these reason, I did not include any transactions in Aberdeenshire in my analysis.


If you have any questions or would like more details about my data decisions, cleaning, or thought-process, please contact me on LinkedIn

Analysis

Prompt #1
Typical time between sales
⚫ Show the median time to the next recorded sale, in years.
⚫ Break it down by property type and postcode area.

• View UK residential turnover rate heat map in Tableau Public

Insights

• View Turnover rate by postcode area graph in Tableau Public

Insights

• View Turnover rate by property ladder graph in Tableau Public

• Click for a glossary

The two biggest challenges with this prompt to show typical times between sales was to calculate the time between sales and to isolate the post code area.I first used Snowflake SQL to join the two tables that were a part of the data set. I selected just the columns that I needed for this prompt. I also sorted by the property ID and date of sale. This was to help calculate the time between sales. I exported this query as a CSV.In Excel, I used a formula to calculate the time between sales in a new column. The formula checked the row above to see if it was the same property. If so, then it subtracted the two sale dates.The postcode area is a part of the postcode, which is a part of an address in the UK. I solved this problem in Excel. First, I used a formula to isolate the postcode fro the rest of the address in a new column. Next, I used a second formula to isolate just the postcode area from the postcode in a new column. This was the trickiest as some area have one letter and some have two.Example 1
Address: 16, Waddon Close, Croydon, Greater London CR0 4JT
Postcode: CR0 4JT
Area: CRExample 2
Address: 15, Apartment 23, Wolstenholme Square, Liverpool, Merseyside L1 4JL
Postcode: L1 4JL
Area: LI exported this CSV with the new columns to Tableau. I also used a pivot table in Excel to explore the data and to confirm what I found using Tableau.My Mistake
When I read the prompt stating, " Break it down by property type and postcode area,' I read it as two different breakdowns. I was wrong. I was being asked to break down the medians using two different variables at the same time. I wished I had asked for clarification earlier.I am glad that I showed my supervisor for this task my work at the midway point. He was able to point out my misunderstanding. I had lots of time to create the desired visualization. I have kept the two graphs that I created to break down the turnover time broken down by postcode area and property type separately.



If you have any questions or would like more details about my analysis, thought-process, code, or formulas, please contact me on LinkedIn

Prompt #2
Price ladder by property type in one area
⚫ Pick one postcode area and one year.
⚫ Show median sale price for Flat, Terraced, Semi detached, Detached.

• View Guildford property ladder graph in Tableau Public

Insights

• Click for a glossary

The first thing I did to answer this prompt was to research what a property ladder is. A property ladder has different types on each rung, with the higher rungs typically being more desirable and most expensive. The bottom rung is thought of as a starter home. In the United Kingdom, the four rungs are from bottom to top are flat, terraced, semi-detached, and detached. For more details, please see the glossary.The biggest challenge with this prompt was picking an area. The first four postcode area I picked did not have a large enough number of sales for each rung on the property ladder in one year. The data set was missing several postcode areas in Scotland, including Glasgow, there area my grandmother was born in.I picked Guildford because it had more than ten sales for each rung in 2024. 2024 was the most recent complete year in the data set. The town of Guildford has some interesting history. It was first mentioned in the writings that have been found in 880 CE. During World War II, 2500 children from London were taken to Guildford. Guildford has some of the most expensive properties in England outside of London. Plus there is a town in the Guildford area named Bagshot. In The Lord of the Rings, Samwise Gamgee and the the Gaffer lived at Number 3 Bagshot Row.The analysis part of this prompt was much simpler because of the work I had already done with prompt 1. I only needed to filter for the Guildford postcode area, GU, and for properties sold in 2024.The sizing of the vizualation was a challenge. I made the graph in Tableau. When the sizing was good in Tableau, the words were too small on this website. Optimizing for Carrd meant the sizing was not good in Tableau. The medium of data communication is important. An analyst may need multiple versions depending on how and where they are communicating.

If you have any questions or would like more details about my analysis, thought-process, code, or formulas, please contact me on LinkedIn

Prompt #3
For each postcode area, compare median sale price across Leasehold and Freehold tenure.

• View median price by tenure table in Tableau Public

• Click for a glossary

This prompt was my favorite because I got to do some research and learn how to do new things in Tableau.The sales prices had a pound symbol in the data. I used Excel to remove it so that I could make calculations with numbers.To answer the prompt, I created a two-way table in Tableau that had showed the median price for each tenure type in each postcode area. I noticed that the median freehold price was greater than the median leasehold price in all areas. This made sense because owners have more freedom with a property that is a freehold.I was curious how much more did a freehold cost in each area.I did not know how to get Tableau to subtract the median leasehold price from the median freehold price in each area. I could not create a calculated field that said free minus lease because the tenure type data was all in the same field.I did some research and was able to solve this problem by using relative columns. I could create a field that would subtract a column by one to its left. I was not satisfied with this approach as the results would change if the columns were moved.I asked several people who have used Tableau more than me for their advice. One was able to show me how to use helper created fields to separate the prices for freehold and leasehold even though they were in the same field.Next, I took the logic of my advisor and created another column to calculate the percent increase from leasehold to freehold.When I sorted the table by the percent increase, I noticed that the largest percent increases were in the areas that had some of the larger median prices in general. Sometimes looking at data can lead to false conclusions. I calculated a linear regression with the median leasehold price for each area and the percent increase to the median freehold price. There was a strong positive correlation. The correlation coefficient (r) was about 0.6.
If you have any questions or would like more details about my analysis, thought-process, code, or formulas, please contact me on LinkedIn

Conclusion and Personal Reflections

I enjoyed this project. I liked how it was tool agnostic. I was able to use a combination of SQL, Excel, and Tableau. I was not given any directions are how to answer the prompts. The approaches I took did not have to be the same as anyone else.I also liked how I had to make decisions about some of the messy data. These were decisions that did not have a simple right or wrong answer.

Recommendation

Location. Location. Location.That is a real estate expression of wisdom, at least in the USA. This project shows that there are more variables to real estate pricing and turnover than location.I recommend that everyday home buyers and sellers find someone who can collect and analyze data with them. Real estate can be complicated. Having information about trends and averages can help to make sure they are getting a deal.

Call to Action

If you're interested in data-driven insights or need a data analyst, I'd love to connect and explore how I can help your business grow. Feel free to reach out!One of my strengths is curiosity and researching skills.I did not know much about real estate in the United Kingdom before this project. I had not done some of the data cleaning that was needed before.Could your organization benefit from my curiosity-driven analyses? Do you like to interact with a lifelong learner who improves his skills with each project?Let’s connect.
Email [email protected]
Connect on LinkedIn

• Home
• Projects
• About Me
• Resume
• LinkedIN

Glossary

Sources

• Home
• Projects
• About Me
• Resume
• LinkedIN

💲 Financial Support for Zimbabwe

Once upon a time, I worked in the grounds and maintenance department for a small graduate school. Most of my coworkers were international students. We became friends, sharing meals after hours.One task that two people had to do each day was to clean bathrooms and pick up trash from offices. I was cleaning a restroom with a colleague I knew was from Zimbabwe. I asked him what town he was from. He said his last name. Whoa. He lived in a village that was named for his family.After I was married, I discovered they had welcomed a student from Zimbabwe as a close family friend.

Why I Chose This Project

I chose this project for my first SQL project for a few reasons. One is because the data set has over one million rows. It has too many rows from Excel. This is the type of data set where I have to use SQL. A second reason is that world economics has been in the news often in the past year. With the United States cutting aid to other countries and increasing tariffs, loans and grants from other institutions, such as the International Development Association (IDA), might be even more important in the near future.I chose to focus on Zimbabwe because of the kind and wise people I have met who live there.

What You'll Learn

You will learn that Zimbabwe could use some help. According to The Atlas for Economic Complexity, Zimbabwe ranks in the bottom 25% for GDP and economic complexity. Yet, it has not been in the top 25% for projects supported by loans or grants from the IDA. Zimbabwe has currently paid back a little more than half of its loans from the IDA.

Key Takeaways

⚫ Zimbabwe ranks 66th out of 144 countries for money loaned or granted by the International Development Association (IDA).
⚫ Zimbabwe ranks 78th out of 147 countries for the number of projects that have received money from the IDA.
⚫ Zimbabwe has repaid 54% of their loans to the IDA.

Dataset Details

• Data source: IDA Statement Of Credits, Grants and Guarantees - Historical Data

The data set comes from the World Bank Bank group. The version I used was up to date as of December 31, 2025. The data has information about loans and grants from the International Development Association (IDA) to countries and other economies around the world. Since 2011, each project has been updated monthly. Old data is kept in the data set after a new update.The data set has over one million (1,000,000) rows. It has 33 columns.

Analysis: Process and Results

Tools

For this project, I mostly used Snowflake SQL. The data set had too many rows to analyze with Excel. I chose Snowflake because I had experience with it.

Data cleaning 1

Since I was answering questions about Zimbabwe, I wanted to make sure that any queries would get all of the data for Zimbabwe. I checked the listed names of all countries that started with Z. I used the SUBSTR function to look at the first letter only in each country name. I used the LOWER function in case there were formatting differences.The data set did not need cleaning. All of the Zimbabwe listings were spelled correctly and all were capitalized.

• Click for CSV of the results of this query.

Data cleaning 2

Next, I wanted to check the formatting of all of the Zimbabwe rows. By ensuring that I could get all of the Zimbabwe data with a simple WHERE command, I would not need to check nearly as many rows.Two columns had inconsistent formatting involving capitalization, “BORROWER” and “PROJECT_NAME.” However, each borrower and each project name was consistent. For example, the borrower “H. M. TREASURY” had all capital letters each time it was used and “Zimbabwe Electricity Supply Commission” had each word capitalized each time it was used. I decided to use the INITCAP function in future queries involving the project name or Borrower to improve consistency.

• Click for CSV of the results of this query.

Question 1. How much money has the IDA loaned or granted to Zimbabwe?

$886,002,975.16 (US$)I used the data dictionary to make sure that the column “Disbursed Amount (US$)” contained the exact amount that had been disbursed for each project. From an earlier query, I knew that each project ID and project name in Zimbabwe was listed once. So, I could get SQL to simply add all of the numbers in the “Disbursed Amount (US$)” column.

• Click for CSV of the results of this query.

Question 2. How many projects has Zimbabwe received from the IDA?

22 projectsSimilar to the previous question, this one is simpler because each row of data involving Zimbabwe has a distinct project. I used the COUNT function to find the total number of projects. I checked loan_number, project_id, and project_name to confirm.

• Click for CSV of the results of this query.

Question 3. Where does Zimbabwe rank globally in terms of number of projects with loans or grants from the IDA?

66th of 147In the previous question, I used SQL to determine that the IDA has given a loan or grant to Zimbabwe for 22 projects. This number needed context. I wrote an SQL query to see how 22 projects compared to the rest of the world.I used DISTINCT when counting the loan numbers because the data after 2011 is a monthly snapshot. Many other countries have the same project listed multiple times. I grouped by country. Then I sorted by the number of projects from most to least. This sorting made it simple to find Zimbabwe and their 22 projects.

• Click for CSV of the results of this query.

Question 4. How much of the money the IDA has extended to Zimbabwe has been paid back?

⚫ $462,773,923.11 (USD)
⚫ 54% repaidThis question is similar to question 1. I had already checked and seen that each Zimbabwe row was a different project. I only needed to add the amounts in the “Repaid to IBRD (US$)” column.

• Click for CSV of the results of this query.

This number alone lacks context. I decided to find the percent repaid. I calculated the total amount that needs to be repaid using the “Due to IBRD (US$)” column. I used the sums from "Due to IBRD (US$)” and “Repaid to IBRD (US$)” to calculate the percent that has been repaid so far.I wasn’t sure that SQL could do calculations like this or what the syntax of a calculation was. I thought I would try it and see what happens. My instinct was correct. Score one for my combination of experiences in studying several coding languages, Excel formulas, and teaching math.

• Click for CSV of the results of this query.

Question 5. What are the names of the top 5 projects the IDA has extended money into Zimbabwe?

⚫ Sap
⚫ Power I
⚫ Power III
⚫ Urban II
⚫ Manu.Sect.Export ProI used the DISTINCT function, though it was not needed in this case. I realize that many analyst’s tasks I will encounter will have much more than 22 rows that Zimbabwe has in this data set. I used the ORDER BY function and DESC to sort the projects from most money to least. I used the LIMIT function to make sure I only reported the top five (5).

• Click for CSV of the results of this query.

Question 6. Which project in Zimbabwe had the greatest amount of money cancelled and how much was cancelled?

⚫ Railway Development
⚫ $17,439,748.42I could have written one query to find the greatest amount cancelled using MAX. I could have taken that amount and used it as a part of the WHERE clause in a second function to find the project name. Since two purposes of doing projects are practicing my skills and showing what I can do, I wanted to get the results with one query.I remembered something about subqueries from tutorials I completed back in five months ago. I did not remember exactly how to do it. First, I wrote down what I wanted the SQL code to do. I wanted the output to be the project name and the amount cancelled only for the project that had the most cancelled. This helped me to realize that the subquery needed to be in the WHERE clause. It took me a few tries to get the syntax right.

• Click for CSV of the results of this query.

Question 7. What is Zimbabwe’s ranking among other countries/economies for most money loaned or granted?

66th of 147The data set was designed so that each row was a snapshot in time. This means that the same project can have more than one row. It turned out that none of the Zimbabwe data had multiple rows for the same project. This is not true for all of the other countries.I was given a block of code that selects only the most recent row for each country/economy. I used that WITH function to help answer this question. After working through other questions in the project, especially question five [5], I understand this block of code much more. I did have to make some changes. The IDA Statement Of Credits, Grants and Guarantees - Historical Data table has changed column names since that code was created.Once I got a temporary table with just the most recent information for each project, I used that in my FROM to make the calculations to answer this question. I knew I wanted a column with the country/economy’s name so that I could easily find Zimbabwe. I also needed the total amount loaned/granted, which I had calculated in a previous question. I used the same formula of “Original Principal Amount (US$)” minus “Cancelled Amount (US$).” I grouped by "Country / Economy" so I could get a total for each nation. I also rounded to two decimal places, standard for US dollars. Then, I sorted so that the highest total loaned/granted was on top.

• Click for CSV of the results of this query.

Question 8. How many projects in Zimbabwe involved transportation?

6 projectsThis question is made easier with Zimbabwe not having many rows in this data set. I took a CSV from earlier that included all of the project names in Zimbabwe. I scanned these for ones involving transportation.Some of the projects had titles that started the same, such as “RAILWAY” and “RAILWAY DEVELOP.” I used the SUBST function in the WHERE clause so that I did not have to enter as much code. I used LOWER because there are different capitalizations in the “PROJECT_NAME” column.

• Click for CSV of the results of this query.

Conclusion and Personal Reflections

Using SQL to analyze this data set increased my confidence that I am a data analyst. This was my first data set with over 1,000,000 rows. The way I approached question 6 was especially helpful to me. By first writing down the output I desired and then determining the code structure, I showed myself that I can take what I have learned throughout my life and apply to new tools. In college, I took classes on multiple programming languages, including C++, Visual BASIC, and COBOL.I can do much more than follow along with tutorials.The results of this project have me wondering why Zimbabwe has received so little from the IDA when other economic stats show that it is not a wealthy nation and could use more support. There are no entries in this data set for Zimbabwe in the past decade.

Recommendation

People who care about Zimbabwe should look into reasons why there has been no activity between Zimbabwe and the IDA for over ten years. The United States has disbursed about 40% less aid in Zimbabwe in fiscal year 2025 than 2024. This is the lowest since 2015 and is not adjusted for inflation. In 2023, the United States was the largest governmental supporter of Zimbabwe (U.S. Agency for International Development and U.S. Department of State, 2025).Folks who care about Zimbabwe, both in and out of the nation, need to get advice from experts in economics and diplomacy to brainstorm how to make up this $150,000,000 difference plus find more to encourage a stronger and more complex economy for the good of the Zimbabwe people.

Call to Action

If you're interested in data-driven insights or need a data analyst, I'd love to connect and explore how I can help your business grow. Feel free to reach out!One of my strengths is that I have been working with computers since the 1980s, often without a lot of resources. I have spent a large portion of my life figuring things out by thinking about output and input, backing up, trying something, and making adjustments using error messages.Could your organization benefit from my curiosity-driven analyses? Do you like to interact with a lifelong learner who first learned to code reading the manual to a BASIC ROM cartridge?Let’s connect.

• LinkedIN Profile

• Home
• Projects
• About Me
• Resume
• LinkedIN

Sources

• Home
• Projects
• About Me
• Resume
• LinkedIN