This project explores patterns in reported UFO shapes and encounter durations. I will identify the three most frequently reported shapes and compare how long their encounters reportedly lasted. These observations describe witness reports and do not establish what the objects actually were.
Research Questions
Which three UFO shapes are reported most frequently?
How do encounter durations compare among these shapes?
Data Overview
The reports originally came from the National UFO Reporting Center and are available through the CORGIS Dataset Project. This file contains 60,632 observations and 16 variables. Each observation represents a reported sighting and includes information about its location, shape, duration, description, and dates. The file also contains U.S. reports with sighting years ranging from 1910 through 2014.
Reported shape is categorical, while encounter duration is quantitative and measured in seconds. Country and state provide categorical location information, and sighting year identifies when the encounter reportedly occurred.
Rows: 60632 Columns: 16
── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr (5): Location.City, Location.State, Location.Country, Data.Shape, Data....
dbl (11): Data.Encounter duration, Location.Coordinates.Latitude, Location.C...
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
2. Exploring the Dataset
Before creating visualizations, I check the dataset’s dimensions, column names, and missing values. I also examine country codes and sighting years to confirm the geographic and historical coverage.
# Display the number of observations and variables and the column names.dim(ufo)
# Check which countries appear and summarize the sighting years.table(ufo$Location.Country)
US
60632
summary(ufo$Dates.Sighted.Year)
Min. 1st Qu. Median Mean 3rd Qu. Max.
1910 2002 2007 2004 2011 2014
3. Counting the Reported Shapes
I count the reports in each shape category and calculate their proportions of the entire dataset. Sorting the results identifies the most frequently reported shapes.
# Count reports for each shape and calculate its share of all reports.# Sort the results from the most common shape to the least common.shape_counts <- ufo %>%count(Data.Shape) %>%mutate(proportion = n /sum(n)) %>%arrange(desc(n))shape_counts
# A tibble: 28 × 3
Data.Shape n proportion
<chr> <int> <dbl>
1 light 12895 0.213
2 triangle 6268 0.103
3 circle 5890 0.0971
4 fireball 4939 0.0815
5 unknown 4359 0.0719
6 other 4209 0.0694
7 sphere 4134 0.0682
8 disk 3853 0.0635
9 oval 2881 0.0475
10 formation 1908 0.0315
# ℹ 18 more rows
4. Selecting the Three Most Common Shapes
Light, triangle, and circle are the three most frequently reported categories. Together, they contain 25,053 reports, approximately 41.3% of the dataset. Both visualizations focus on these categories.
# Keep the three most common reported shapes.top_shapes <- ufo %>%filter(Data.Shape %in%c("light", "triangle", "circle"))# Display the selected reports.top_shapes
# A tibble: 25,053 × 16
Location.City Location.State Location.Country Data.Shape
<chr> <chr> <chr> <chr>
1 anchorage AK US circle
2 anchorage AK US circle
3 anchorage AK US circle
4 anchorage AK US circle
5 anchorage AK US circle
6 anchorage AK US circle
7 anchorage AK US circle
8 anchorage AK US circle
9 anchorage AK US circle
10 anchorage AK US circle
# ℹ 25,043 more rows
# ℹ 12 more variables: `Data.Encounter duration` <dbl>,
# `Data.Description excerpt` <chr>, Location.Coordinates.Latitude <dbl>,
# Location.Coordinates.Longitude <dbl>, Dates.Sighted.Year <dbl>,
# Dates.Sighted.Month <dbl>, Date.Sighted.Day <dbl>,
# Dates.Sighted.Hour <dbl>, Dates.Sighted.Minute <dbl>,
# Dates.Documented.Year <dbl>, Dates.Documented.Month <dbl>, …
5. Comparing the Number of Reports
A bar chart compares the frequencies of the three selected categories. The colors distinguish the shapes, and the legend identifies each category.
# Count and graph the reports for each selected shape.ggplot(top_shapes, aes(x = Data.Shape, fill = Data.Shape)) +geom_bar() +scale_fill_manual(values =c(circle ="#0072B2",light ="#D55E00",triangle ="#009E73" ) ) +theme_bw() +labs(title ="Three Most Common Reported UFO Shapes",x ="Reported UFO Shape",y ="Number of Reported Sightings",fill ="Reported Shape",caption ="Source: NUFORC reports via CORGIS UFO Sightings CSV" )
Light is the most frequently reported category, with 12,895 reports, followed by triangle with 6,268 and circle with 5,890. Light reports are more than twice as common as either triangle or circle reports. The difference between triangle and circle is considerably smaller. These counts describe reports in this collection rather than the actual frequency of unidentified objects.
6. Preparing Encounter Durations
For the duration comparison, I keep positive, nonmissing encounter durations. Positive values are necessary for a logarithmic axis. I retain unusually long encounters so their influence remains visible.
# Keep positive durations that are not missing.duration_data <- top_shapes %>%filter(!is.na(`Data.Encounter duration`) &`Data.Encounter duration`>0 )# Display the filtered reports.duration_data
# A tibble: 25,053 × 16
Location.City Location.State Location.Country Data.Shape
<chr> <chr> <chr> <chr>
1 anchorage AK US circle
2 anchorage AK US circle
3 anchorage AK US circle
4 anchorage AK US circle
5 anchorage AK US circle
6 anchorage AK US circle
7 anchorage AK US circle
8 anchorage AK US circle
9 anchorage AK US circle
10 anchorage AK US circle
# ℹ 25,043 more rows
# ℹ 12 more variables: `Data.Encounter duration` <dbl>,
# `Data.Description excerpt` <chr>, Location.Coordinates.Latitude <dbl>,
# Location.Coordinates.Longitude <dbl>, Dates.Sighted.Year <dbl>,
# Dates.Sighted.Month <dbl>, Date.Sighted.Day <dbl>,
# Dates.Sighted.Hour <dbl>, Dates.Sighted.Minute <dbl>,
# Dates.Documented.Year <dbl>, Dates.Documented.Month <dbl>, …
7. Summarizing Encounter Durations
I calculate the number of reports, average duration, and standard deviation for each category. These statistics help describe the differences and variation in reported durations.
# Group reports by shape.# Calculate the report count, average duration, and standard deviation.duration_summary <- duration_data %>%group_by(Data.Shape) %>%summarise(reports =n(),mean_seconds =mean(`Data.Encounter duration`),sd_seconds =sd(`Data.Encounter duration`),.groups ="drop" )duration_summary
The average reported duration is approximately 15,662 seconds for light, 1,439 seconds for triangle, and 2,746 seconds for circle. Extremely long reports strongly influence these averages. The large standard deviations also indicate substantial variation within the categories. The box plots provide another way to compare their typical durations.
8. Comparing Encounter Durations
Side-by-side box plots compare encounter durations across the three shapes. A logarithmic axis displays both short and extremely long encounters on the same graph. Equal distances on this axis represent equal ratios rather than equal differences in seconds.
# Compare the duration distributions using a box plot for each shape.# Use a logarithmic axis because reported durations span a very wide range.ggplot( duration_data,aes(x = Data.Shape, y =`Data.Encounter duration`, fill = Data.Shape)) +geom_boxplot() +scale_y_log10() +scale_fill_manual(values =c(circle ="#0072B2",light ="#D55E00",triangle ="#009E73" ) ) +theme_bw() +labs(title ="Reported UFO Encounter Durations by Shape",x ="Reported UFO Shape",y ="Encounter Duration (seconds; log10 scale)",fill ="Reported Shape",caption ="Source: NUFORC reports via CORGIS; positive durations only" )
The three categories have similar median durations of approximately 180 seconds, or three minutes. Their distributions overlap despite the differences between their averages. Each category contains unusually short and long encounters. Points beyond the whiskers indicate unusual durations under the box plot’s rule, but they are not automatically recording errors.
Conclusion
Light reports are considerably more frequent than triangle or circle reports in this dataset. However, the three categories have similar typical encounter durations, with medians near three minutes. The differences between their averages and medians show how extreme values can influence an average. Examining both frequencies and duration distributions provides a clearer understanding of these reports.
