provide a list of medium sized R databases that will be used to perform a regression of…
- User
provide a list of medium sized R databases that will be used to perform a regression of a predicted variable consisting of counts, and predictors that are NOT categorical
Assistant
When performing a regression where the response variable consists of counts, and the predictors are not categorical, you're likely looking at Poisson regression, negative binomial regression, or other count-based models. In R, there are several built-in and contributed datasets that can be used for such purposes.
Below is a list of medium-sized datasets in R (base or commonly-used packages) that are suitable for regression with count response variables and continuous predictors:
🔹 1.
quakes(base R)- Description: Locations of 1000 seismic events (earthquakes) near Fiji.
- Response (Y):
depthorstations(could be treated as count). - Predictors (X): Latitude, longitude (continuous), magnitude.
- Use case: Predict number of stations recording an earthquake (
stations) based on location and magnitude. - Size: 1000 rows, 5 columns
data(quakes) head(quakes)
🔹 2.
discoveries(base R)- Description: The number of great inventions and scientific discoveries in each year from 1860 to 1959.
- Response (Y):
discoveries(count) - Predictors (X): Year (can be transformed into continuous time variable)
- Use case: Time series regression of discovery counts over time.
- Size: 100 rows, 1 column
data(discoveries) plot(discoveries)
🔹 3.
Insurance(MASS package)- Description: Data from a Swedish insurance company showing the number of claims per policyholder.
- Response (Y):
Claims(count) - Predictors (X): Kilometres driven, Bonus, Make (but can filter/remove categorical variables)
- Use case: Model number of insurance claims using continuous predictors like distance driven.
- Size: 63 rows, 5 columns
library(MASS) data(Insurance) head(Insurance)
🔹 4.
bioChemists(pscl package)- Description: Number of articles published by graduate students in biochemistry.
- Response (Y):
art(number of articles published – count) - Predictors (X):
ment(number of articles mentored – continuous),phd,mar, etc. - Use case: Poisson or negative binomial regression predicting publication count.
- Size: 915 rows, 7 columns
library(pscl) data(bioChemists) head(bioChemists)
🔹 5.
RecreationDemand(AER package)- Description: Recreation demand at a lake; includes number of trips taken.
- Response (Y):
trips(count) - Predictors (X): Cost variables (e.g.,
income,cost,out_of_pocket_cost) - Use case: Model recreation demand (count data) using economic and demographic factors.
- Size: 659 rows, 7 columns
library(AER) data(RecreationDemand) head(RecreationDemand)
🔹 6.
CrabSatellites(MASS package)- Description: Number of satellites (male crabs) around female horseshoe crabs.
- Response (Y):
satellites(count) - Predictors (X): Size, weight (continuous), color (categorical – can exclude)
- Use case: Count regression predicting satellite presence.
- Size: 173 rows, 8 columns
library(MASS) data(crabs) CrabSatellites <- crabs head(CrabSatellites)
✅ Summary Table
Dataset Package Response Continuous Predictors Rows Notes quakesbase stationslat, long, mag 1000 Location-based prediction discoveriesbase discoveriesyear 100 Time-series count modeling InsuranceMASS ClaimsKilometres, Bonus 63 Insurance claim modeling bioChemistspscl artment,phd,mar915 Publication count analysis RecreationDemandAER tripscost, income 659 Economic modeling CrabSatellitesMASS satellitessize, weight 173 Ecological / biological modeling
Let me know if you'd like code examples for fitting a Poisson or negative binomial model on any of these datasets!