For my response variable, I will be choosing “Units Sold”. Book units sold is of great interest to both authors and publishers as this is a direct reflection of how successful a book is. If a book is in high demand, it is certain to be successful, regardless of how people actually feel about the book.
The categorical variable that I will choose is publisher. I chose publisher because I wish to see if different publishers have any sort of impact on the number of book units sold.
I hope to examine if there is enough data to conclude whether or not average units sold per book for each publisher varies greatly. This analysis could supply an author with the knowledge of whether or not it is worth to sign with one publisher over another.
My Null Hypothesis is:
\[ H_0 : \text{The average units sold is the same across all publishers.} \]
#performing anova
anova_result <- aov(`units sold` ~ Publisher, data = book_sales)
summary(anova_result)## Df Sum Sq Mean Sq F value Pr(>F)
## Publisher 6 8.589e+08 143152133 0.605 0.727
## Residuals 1063 2.517e+11 236779969Operating on a significance level of 0.05, a P-Value of .727 means that there is not enough evidence for the null hypothesis (that the average units sold is the same across all publishers) to be rejected. Based on this, it is possible to assume that the choice of publisher does not seem to be a significant factor influencing the number of units sold. There is not enough evidence to conclude that there is a difference in the average units sold across each publisher, so it would be reasonable to assume that the publisher does not play a large role in determining the number of book units sold.
In order to investigate another potential factor that could influence units sold, I will be fitting a linear regression model. I chose sale price to investigate because this is a very common consideration that consumers take into account when participating in any sort of retail. By investigation this, I would like see how the sales price of a book effects how many units of that book are sold.
#linear regression model
lm_sale_price <- lm(`units sold` ~ `sale price`, data = book_sales)
summary(lm_sale_price)##
## Call:
## lm(formula = `units sold` ~ `sale price`, data = book_sales)
##
## Residuals:
## Min 1Q Median 3Q Max
## -9904 -9116 -5788 -4353 51771
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 9587.66 796.81 12.033 <2e-16 ***
## `sale price` 18.34 132.12 0.139 0.89
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15380 on 1068 degrees of freedom
## Multiple R-squared: 1.805e-05, Adjusted R-squared: -0.0009183
## F-statistic: 0.01928 on 1 and 1068 DF, p-value: 0.8896The coefficient of 18.34 means that for each additional unit increase in the book’s sale price, the model predicts an increase of 18.34 units sold, however, the p-value for this coefficient is 0.89. This p-value is greater than the standard significance level of 0.05, therefore it is reasonable to assume that there is not enough evidence to believe that sales price effects the units sold.
In the context of my data, I believe further exploration and consideration of other factors (such as genre or author rating) might provide a better understanding of any other factors that may influence how many units of a book are sold.