R Notebook By Bhavya Kalra
Beer Profile and Ratings Analysis
About the Data:
The Beer Profile and Ratings dataset from Kaggle was used for the project. The main data set (beer_profile_and_ratings.csv) contains the following columns: (General) • Name: Beer name (label) • Style: Beer Style • Brewery: Brewery name • Beer Name: Complete beer name (Brewery + Brew Name) • Description: Notes on the beer if available • ABV: Alcohol content of beer (% by volume) • Min IBU: The minimum IBU value each beer can possess • Max IBU: The maximum IBU value each beer can possess
(Mouth feel) • Astringency • Body • Alcohol (Taste) • Bitter • Sweet •Sour • Salty (Flavor And Aroma) • Fruits • Hoppy • Spices • Malty
(Reviews) • review_aroma • review_appearance • review_palate •review_taste • review_overall • number_of_reviews
Loading the libraries
library(tidyverse)## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.2 ✔ readr 2.1.4
## ✔ forcats 1.0.0 ✔ stringr 1.5.0
## ✔ ggplot2 3.4.3 ✔ tibble 3.2.1
## ✔ lubridate 1.9.2 ✔ tidyr 1.3.0
## ✔ purrr 1.0.2
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorslibrary(ggplot2)
library(reshape2)##
## Attaching package: 'reshape2'
##
## The following object is masked from 'package:tidyr':
##
## smithslibrary(dplyr)
library(gridExtra)##
## Attaching package: 'gridExtra'
##
## The following object is masked from 'package:dplyr':
##
## combinelibrary(boot)
library(pwr)# Loading the dataset in beers data frame
beers <- read.csv("/Users/bhavyakalra/Desktop/Stats R/Final_project_beer/beer_profile_and_ratings.csv")
head(beers)## Name Style
## 1 Amber Altbier
## 2 Double Bag Altbier
## 3 Long Trail Ale Altbier
## 4 Doppelsticke Altbier
## 5 Sleigh'r Dark Doüble Alt Ale Altbier
## 6 Sticke Altbier
## Brewery
## 1 Alaskan Brewing Co.
## 2 Long Trail Brewing Co.
## 3 Long Trail Brewing Co.
## 4 Uerige Obergärige Hausbrauerei GmbH / Zum Uerige
## 5 Ninkasi Brewing Company
## 6 Uerige Obergärige Hausbrauerei GmbH / Zum Uerige
## Beer.Name..Full.
## 1 Alaskan Brewing Co. Alaskan Amber
## 2 Long Trail Brewing Co. Double Bag
## 3 Long Trail Brewing Co. Long Trail Ale
## 4 Uerige Obergärige Hausbrauerei GmbH / Zum Uerige Uerige Doppelsticke
## 5 Ninkasi Brewing Company Sleigh'r Dark Doüble Alt Ale
## 6 Uerige Obergärige Hausbrauerei GmbH / Zum Uerige Uerige Sticke
## Description
## 1 Notes:Richly malty and long on the palate, with just enough hop backing to make this beautiful amber colored "alt" style beer notably well balanced.\\t
## 2 Notes:This malty, full-bodied double alt is also known as “Stickebier” – German slang for “secret brew”. Long Trail Double Bag was originally offered only in our brewery taproom as a special treat to our visitors. With an alcohol content of 7.2%, please indulge in moderation. The Long Trail Brewing Company is proud to have Double Bag named Malt Advocate’s “Beer of the Year” in 2001. Malt Advocate is a national magazine devoted to “expanding the boundaries of fine drinks”. Their panel of judges likes to keep things simple, and therefore of thousands of eligible competitors they award only two categories: “Imported” and “Domestic”. It is a great honor to receive this recognition.33 IBU\\t
## 3 Notes:Long Trail Ale is a full-bodied amber ale modeled after the “Alt-biers” of Düsseldorf, Germany. Our top fermenting yeast and cold finishing temperature result in a complex, yet clean, full flavor. Originally introduced in November of 1989, Long Trail Ale beer quickly became, and remains, the largest selling craft-brew in Vermont. It is a multiple medal winner at the Great American Beer Festival.25 IBU\\t
## 4 Notes:
## 5 Notes:Called 'Dark Double Alt' on the label.Seize the season with Sleigh'r. Layers of deeply toasted malt are balanced by just enough hop bitterness to make it deceivingly drinkable. Paired with a dry finish, Sleigh’r is anything but your typical winter brew.An Alt ferments with Ale yeast at colder lagering temperatures. This effect gives Alts a more refined, crisp lager-like flavor than traditional ales. The Alt has been “Ninkasified” raising the ABV and IBUs. Sleigh'r has a deep, toasted malt flavor that finishes dry and balanced.50 IBU\\t
## 6 Notes:
## ABV Min.IBU Max.IBU Astringency Body Alcohol Bitter Sweet Sour Salty Fruits
## 1 5.3 25 50 13 32 9 47 74 33 0 33
## 2 7.2 25 50 12 57 18 33 55 16 0 24
## 3 5.0 25 50 14 37 6 42 43 11 0 10
## 4 8.5 25 50 13 55 31 47 101 18 1 49
## 5 7.2 25 50 25 51 26 44 45 9 1 11
## 6 6.0 25 50 22 45 13 46 62 25 1 34
## Hoppy Spices Malty review_aroma review_appearance review_palate review_taste
## 1 57 8 111 3.498994 3.636821 3.556338 3.643863
## 2 35 12 84 3.798337 3.846154 3.904366 4.024948
## 3 54 4 62 3.409814 3.667109 3.600796 3.631300
## 4 40 16 119 4.148098 4.033967 4.150815 4.205163
## 5 51 20 95 3.625000 3.973958 3.734375 3.765625
## 6 60 4 103 4.007937 4.007937 4.087302 4.192063
## review_overall number_of_reviews
## 1 3.847082 497
## 2 4.034304 481
## 3 3.830239 377
## 4 4.005435 368
## 5 3.817708 96
## 6 4.230159 315Null Hypothesis 1 (Hypothesis related to ABV):
Null Hypothesis (H0): The average ABV (Alcohol by Volume) for Altbier-style of beers is equal to 5.3%.
Alternative Hypothesis (Ha): The average ABV for Altbier-style beers is not equal to 5.3%.
Alpha Level (α): 0.05 (This is a common choice for alpha in hypothesis testing). Power Level (1 - β): 0.80 (A power level of 0.80 is often used to ensure a reasonable chance of detecting an effect if it exists). Minimum Effect Size (δ): 0.2 (This value represents the minimum meaningful difference in ABV that we want to detect). We can perform a two-sample t-test to compare the average ABV of Style beers in the dataset to the hypothesized value of 5.3%. We’ll calculate the test statistic and p-value and then determine if we have enough data to perform a Neyman-Pearson hypothesis test.
# Hypothesis 1: ABV Test
hypothesized_abv <- 5.3
alpha <- 0.05
power <- 0.80
min_effect_size <- 0.2
abv_test <- t.test(beers$ABV, mu = hypothesized_abv, alternative = "two.sided")
effect_size_abv <- abs((mean(beers$ABV) - hypothesized_abv) / sd(beers$ABV))
required_sample_size_abv <- pwr.t.test(d = effect_size_abv, sig.level = alpha, power = 0.80, type = "two.sample")$n
cat("ABV Test Results:\n")## ABV Test Results:cat("Test Statistic:", abv_test$statistic, "\n")## Test Statistic: 27.23183cat("P-value:", abv_test$p.value, "\n")## P-value: 4.987927e-147cat("Effect Size:", effect_size_abv, "\n")## Effect Size: 0.4816211cat("Power:", required_sample_size_abv)## Power: 68.64828The ABV test I conducted has already given me compelling evidence against the null hypothesis. The results showed an extremely low p-value and a substantial effect size, indicating a significant difference from the hypothesized value. In light of these findings, I believe that performing a Neyman-Pearson hypothesis test may not be necessary. Typically, Neyman-Pearson testing is used for planning studies and assessing statistical power, but in this case, we already have strong evidence against the null hypothesis. However, it’s important to note that the test’s power is slightly below the desired level, indicating a possibility of not detecting the effect size if it were smaller.
ggplot(beers, aes(x = "Actual ABV vs. Hypothesized ABV", y = ABV, fill = "ABV")) +
geom_boxplot() +
geom_hline(yintercept = 5.3, color = "red", linetype = "dashed") +
labs(title = "Box Plot of ABV for Style Beers",
x = "Test Type",
y = "ABV (Alcohol by Volume)") +
theme_minimal()
This visual observation aligns with the statistical analysis and p-value obtained from the t-test, which also suggested a significant difference in ABV. Therefore, the boxplot provides a graphical representation that supports the statistical conclusion that the average ABV differs from the hypothesized value.
Null Hypothesis 2 (Hypothesis related to Review Scores of Beers):
Null Hypothesis (H0): There is no significant difference in the average overall review scores between beers produced by “Alaskan Brewing Co.” and “Long Trail Brewing Co.”
Alternative Hypothesis (Ha): There is a significant difference in the average overall review scores.
Alpha Level (α): 0.05 (Same as in the previous hypothesis). Power Level (1 - β): 0.85 (I want a higher power level in this case to be more confident in our test results). Minimum Effect Size (δ): 0.2 (This represents the minimum meaningful difference in review scores).
# Hypothesis 2: Review Test
alpha <- 0.05
power <- 0.85
min_effect_size <- 0.2
group1 <- beers$review_overall[beers$Brewery == "Alaskan Brewing Co."]
group2 <- beers$review_overall[beers$Brewery == "Long Trail Brewing Co."]
t_result <- t.test(group1, group2)
t_result##
## Welch Two Sample t-test
##
## data: group1 and group2
## t = 0.76655, df = 9.6916, p-value = 0.4616
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.2073954 0.4235124
## sample estimates:
## mean of x mean of y
## 3.845719 3.737660The results of the statistical analysis, including a p-value of 0.4616 greater than the preset alpha level of 0.05 and a confidence interval spanning from -0.2074 to 0.4235, demonstrate a consistent outcome. This outcome leads to the failure to reject the null hypothesis, indicating that there is insufficient evidence to support a significant difference in the average overall review scores between “Alaskan Brewing Co.” and “Long Trail Brewing Co.”
data_to_plot <- data.frame(
Brewery = factor(c("Alaskan Brewing Co.", "Long Trail Brewing Co.")),
Overall_Review_Score = c(group1, group2)
)
ggplot(data_to_plot, aes(x = Brewery, y = Overall_Review_Score, fill = Brewery)) +
geom_boxplot() +
labs(
title = "Comparison of Overall Review Scores between Breweries",
x = "Brewery",
y = "Overall Review Score"
) +
theme_minimal()
The boxplot comparing the average overall review scores between “Alaskan Brewing Co.” and “Long Trail Brewing Co.” reveals that there is substantial overlap between the two breweries’ distributions. Both breweries have almost similar median review scores. This visual observation aligns with the statistical analysis, which found no strong evidence to reject the null hypothesis.
My dataset does not have enough categorical data columns, the columns with categorical values have large number of cells/data to compute the Contigency table. Therefore, I have not performed the fisher’s test.
# Performing Chi squared test
contingency_table <- table(beers$Style, beers$Brewery)
chi_squared_result <- chisq.test(contingency_table, correct = TRUE)## Warning in chisq.test(contingency_table, correct = TRUE): Chi-squared
## approximation may be incorrectprint(chi_squared_result)##
## Pearson's Chi-squared test
##
## data: contingency_table
## X-squared = 127134, df = 102630, p-value < 2.2e-16