Statistical test

Hypothesis Testing

T-test

Analysis of Variance

To test whether there is a significant difference in means across categorical independent variable(s) on a continuous dependent variable.

One-way ANOVA

Conditions

  • One categorical independent variable with three or more groups

  • One continuous dependent variable

We want to test whether the mean number of bird collisions significantly differs across four orientation groups (E, N, S, W).

Orientation Total Collisions
East (E) 14
North (N) 28
South (S) 35
West (W) 26

Hypotheses

𝐻0 : 𝜇1 = 𝜇2 = 𝜇3 = … = 𝜇𝑘
𝐻1 : At least one mean is different from others.

Steps

  1. Calculate the within-group sum of squares (𝑆𝑆𝑊)
  2. Calculate the between-group sum of squares (𝑆𝑆𝐵)
  3. SST = SSW + SSB
  4. Calculate degrees of freedom (𝑑𝑓)
  5. Calculate mean squares (𝑀𝑆) = SS/df
  6. Calculate F-ratio
  7. Compare F-ratio to critical F-value Suppose we have 𝑘 levels. The i-th level has 𝑛𝑖 observations.
https://study.com/skill/learn/how-to-calculate-the-total-sum-of-squares-within-and-between-ssw-and-ssb-explanation.html

Example with R

library(readxl)

# Load data
data <- read_excel("Bird collision XJTLU_20250601.xlsx", sheet = "Campus")
data_clean <- subset(data, !is.na(`Orientation`) & !is.na(`Bird collision?`))

# Prepare data
data_clean$Orientation <- as.factor(data_clean$Orientation)
data_clean$Collision <- as.numeric(data_clean$`Bird collision?`)

# Check group means
aggregate(Collision ~ Orientation, data = data_clean, mean)
  Orientation Collision
1           E 0.8750000
2           N 0.9655172
3           S 1.0000000
4           W 0.9629630
# Fit one-way ANOVA model
model <- aov(Collision ~ Orientation, data = data_clean)
summary(model)
             Df Sum Sq Mean Sq F value Pr(>F)
Orientation   3  0.172 0.05733   1.605  0.193
Residuals   103  3.678 0.03571               

Interpretation: p<0.05, fail to reject H0

No statistically significant difference in mean bird collisions between the four orientations

Two-way ANOVA

MANOVA

Assumption Test

Post-hoc

Analysis of Covariance

Linear Regression

ANCOVA

MANCOVA

Assumption Test

Post-hoc