quasibin: quasibin: Quasi-binomial Data

In thoree/stat340: Data and functions for BIAS at NMBU

quasibin: Quasi-binomial Data

Description

The quasibin data consists of one binary and three continuous response variables. There are 23 cases of success and 15 cases of no success.

Usage

quasibin

Format

A data frame with 38 observations (rows) and 4 variables (columns).

	Column name	Data type	Description	Values
[,1]	Success	integer	binary response	(0, 1)
[,2]	x1	integer	continuous response	(247 - 3296)
[,3]	x2	numeric	continuous response	(1.2 - 6.1)
[,4]	x3	numeric	continuous response	(0.7 - 14.8)

Details

The quasi-binomial model is in the family of generalized linear models, and can be used in Logistic Regression when the response is binary and one or more assumptions of the binomial distribution are broken.

The variables from the original data set are renamed since the data is not open.

Examples

# First, a look at the data
table(quasibin$Success)
boxplot(quasibin$x3 ~ quasibin$Success)

# Logistic regression
logit1 <- glm(Success ~ x3, data = quasibin,
              family = "binomial")
summary(logit1)

# Surprisingly, `x3` is not a significant predictor.
# A probable reason for this is that observation 7
# is an outlier as seen from:
plot(logit1, which = 1)

# We model dispersion, more reasonable results,
# p-value = 4.57e-05:
logit2 <- glm(Success ~ x3, data = quasibin,
              family = "quasibinomial")
summary(logit2)

thoree/stat340 documentation built on June 30, 2024, 4:04 p.m.