2023-03-26
This data provides a table of popular video games (with over 100,000 sold copies.) The following parameters are covered:
The data set and its details can be found at this page: https://www.kaggle.com/datasets/gregorut/videogamesales
sales = read.csv("vgsales.csv")
colnames(sales)
## [1] "Rank" "Name" "Platform" "Year" "Genre" ## [6] "Publisher" "NA_Sales" "EU_Sales" "JP_Sales" "Other_Sales" ## [11] "Global_Sales"
head(sales)
## Rank Name Platform Year Genre Publisher NA_Sales ## 1 1 Wii Sports Wii 2006 Sports Nintendo 41.49 ## 2 2 Super Mario Bros. NES 1985 Platform Nintendo 29.08 ## 3 3 Mario Kart Wii Wii 2008 Racing Nintendo 15.85 ## 4 4 Wii Sports Resort Wii 2009 Sports Nintendo 15.75 ## 5 5 Pokemon Red/Pokemon Blue GB 1996 Role-Playing Nintendo 11.27 ## 6 6 Tetris GB 1989 Puzzle Nintendo 23.20 ## EU_Sales JP_Sales Other_Sales Global_Sales ## 1 29.02 3.77 8.46 82.74 ## 2 3.58 6.81 0.77 40.24 ## 3 12.88 3.79 3.31 35.82 ## 4 11.01 3.28 2.96 33.00 ## 5 8.89 10.22 1.00 31.37 ## 6 2.26 4.22 0.58 30.26
What are the top 5 most popular publishers? How many games were produced by the rest?
# count the number of games produced per publisher and sort it in decreasing order
publisher = as.data.frame(table(sales$Publisher), stringsAsFactors=FALSE)
colnames(publisher) = c("Publisher", "Frequency")
sort_publisher = publisher[order(publisher$Frequency, decreasing = TRUE),]
rownames(sort_publisher) = c(1:nrow(publisher))
# save the top 5 publishers and group the rest
n = 5
bottom = sort_publisher[(n+1):nrow(publisher),]
other = sum(bottom$Frequency)
publisher_final = sort_publisher[1:n,]
publisher_final[(n+1),] = c("Other", other)
publisher_final
## Publisher Frequency ## 1 Electronic Arts 1351 ## 2 Activision 975 ## 3 Namco Bandai Games 932 ## 4 Ubisoft 921 ## 5 Konami Digital Entertainment 832 ## 6 Other 11587
The top 5 gaming publishers for popular games purchased by consumers were Electronic Arts (1351 games), Activision (975), Namco Bandai Games (932), Ubisoft (921), and Konami Digital Entertainment (832). Other gaming companies produced 11587 games in total.
Many would argue that the success of a gaming company is better marked by the success of the produced games rather than the number of games produced. The box plot shows the success of the top 5 publishers (by number of games produced).
The data can be filtered to better understand a single companies success by country, platform, and relative to other companies. For this project, THQ will be selected as the practice company.
Where does THQ rank by the number of games produced?
index = which(sort_publisher$Publisher == "THQ") index
## [1] 6
sort_publisher[index,2]
## [1] 715
THQ ranks 6th in companies that sell the greatest number of popular video games, with 715 games produced.
Using the data from other companies, THQ could discover which platforms it is most popular with to determine whether it should focus on certain platforms or try to catch up on others. The following slides use data for ALL companies.
# count the number of games produced on each platform and sort them by platform
platforms = as.data.frame(table(sales$Platform), stringsAsFactors=FALSE)
colnames(platforms) = c("Platform", "Frequency")
sort_platforms = platforms[order(platforms$Frequency, decreasing = TRUE),]
rownames(sort_platforms) = c(1:nrow(platforms))
# save the top 5 platforms and group the rest
n = 5
bottom = sort_platforms[(n+1):nrow(platforms),]
other = sum(bottom$Frequency)
platforms_final = sort_platforms[1:n,]
platforms_final[(n+1),] = c("Other", other)
platforms_final
## Platform Frequency ## 1 DS 2163 ## 2 PS2 2161 ## 3 PS3 1329 ## 4 Wii 1325 ## 5 X360 1265 ## 6 Other 8355
The top 5 gaming platforms for popular games purchased by consumers were DS (2163 games), PS2 (2161), PS3 (1329), Wii (1325), and X360 (1265). Other gaming platforms had 8355 games in total.
To be more specific towards THQ purposes, the data will be filtered to show the most popular platforms for THQ games.
# filter the platform data for THQ and recalculate the values
THQ = filter(sales, Publisher == "THQ")
platforms_THQ = as.data.frame(table(THQ$Platform), stringsAsFactors = FALSE)
colnames(platforms_THQ) = c("Platform", "Frequency")
sort_platforms_THQ = platforms_THQ[order(platforms_THQ$Frequency, decreasing = TRUE),]
rownames(sort_platforms_THQ) = c(1:nrow(platforms_THQ))
# save the top 5 platforms and group the rest
n = 5
bottom = sort_platforms_THQ[(n+1):nrow(sort_platforms_THQ),]
other = sum(bottom$Frequency)
platforms_final_THQ = sort_platforms_THQ[1:n,]
platforms_final_THQ[(n+1),] = c("Other", other)
platforms_final_THQ
## Platform Frequency ## 1 DS 115 ## 2 GBA 110 ## 3 PS2 100 ## 4 Wii 76 ## 5 X360 64 ## 6 Other 250
## Analyzing the results By filtering the data to THQ produced-games, a new platform, GBA, is shown with greater significance to the company.
Data analysis can be very valuable to those with two specific interests: video games and the stock market.
Tracking the production of a company over time and its success can help an individual decide whether they’d like to invest in it.
# calculate and plot the line of best fit line_THQ = lm(THQ$Global_Sales~THQ$Year) summary(line_THQ)
## ## Call: ## lm(formula = THQ$Global_Sales ~ THQ$Year) ## ## Residuals: ## Min 1Q Median 3Q Max ## -0.61909 -0.33942 -0.17948 0.08632 2.96080 ## ## Coefficients: ## Estimate Std. Error t value Pr(>|t|) ## (Intercept) 47.212233 12.131227 3.892 0.000109 *** ## THQ$Year -0.023297 0.006047 -3.852 0.000128 *** ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## Residual standard error: 0.5689 on 710 degrees of freedom ## Multiple R-squared: 0.02047, Adjusted R-squared: 0.0191 ## F-statistic: 14.84 on 1 and 710 DF, p-value: 0.0001276
The formula is as follows:
\(Global\;Sales\;(in\;millions) = -0.023297*Year + 47.2122\)
The \(R^2\) value is 0.02047, which suggests that there is a very weak linear correlation between time and sales. There is a lot of variability above and below the plot line. Log-transforming the data did not provide stronger linear trends.
From this data, the company appears to be holding a steady production of games with a slight decline; however, the weak \(R^2\) does not provide confidence to this conclusion.
A second linear interpretation can be performed to see if THQ sales in the United States are indicative of their sales in the rest of the world using a simple linear regression model.
THQ_line = lm(THQ$Global_Sales~THQ$NA_Sales) summary(THQ_line)
## ## Call: ## lm(formula = THQ$Global_Sales ~ THQ$NA_Sales) ## ## Residuals: ## Min 1Q Median 3Q Max ## -0.90914 -0.07868 -0.03032 0.04154 1.17178 ## ## Coefficients: ## Estimate Std. Error t value Pr(>|t|) ## (Intercept) 0.007986 0.010122 0.789 0.43 ## THQ$NA_Sales 1.604765 0.022760 70.508 <2e-16 *** ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## Residual standard error: 0.2032 on 710 degrees of freedom ## Multiple R-squared: 0.875, Adjusted R-squared: 0.8749 ## F-statistic: 4971 on 1 and 710 DF, p-value: < 2.2e-16
The formula is as follows:
\(Global\;Sales= 1.60476*N.A.\;Sales\; + 0.007986\)
with sales in millions.
The \(R^2\) value is 0.875, which suggests that there is a strong linear correlation between sale performance in each grouping. The p-value is also \(2.2*10^{-6}\), so we fail to reject the null hypothesis that there is a relationship between a video game’s sale in North America and the world.
# collect residual values res_THQ = residuals(THQ_line) # create Normal Q-Q plot of residual values qqnorm(res_THQ)
The points curve towards the ends and have a linear trend in the middle, so other than the extremities, the data has a strong linear nature. Other than the very high and very low sales, THQ sales in North America and the world can be reasonably predicted with a linear model.