Indie Author Insights: Word Count While Studying

On this Medium blog, I plan to use statistics and programming to provide insights for indie authors — some general and some related to my personal career. So, who am I? I’ve been an independent author of romance novels since 2013. My undergrad was in environmental studies and English rhetoric and professional writing. In September 2020, I returned to school for a graduate certificate in business insights and analytics. I’ve been learning about statistics, coding, and databases, as well as the business environment, big data, accounting, and marketing. In this post, I’ll use Microsoft Excel and IBM SPSS, and I’ll primarily be referring to Julie Pallant’s SPSS Survival Manual as a resource.

Today, I’m going to look into my word count over the past few years. As an independent author, it’s important to maintain productivity. Self-publishing is like a treadmill where you have to keep constantly producing more books or your earnings will fall off. I’ve been tracking my daily word count for a number of years in spreadsheets like this one.

That makes it easy to come up with monthly and yearly data.

As you can see, it looks like 2021 is pretty low. Of course, we’re only four months in. I did go back to school full-time (six classes per semester!), so that would explain a drop. But I also had zero life because of COVID. I’ve been writing seven days a week instead of five, and even if I only write less per day, it still has to add up — doesn’t it?

Breaking this data down by school year, it’s clear that this past school year was my lowest in terms of output — but there’s a pretty wide range, and I think the standard deviation is pretty large. Maybe the difference isn’t statistically significant.

I’ll give you the answer here in case you’re not into statistics. The only statistically significant difference was between year 1 and year 5. My word count didn’t differ significantly in year 3, 4, and 5. Another cool thing to note is that when I do this analysis by calendar year, including all twelve months of data instead of just eight, there’s actually no significant difference!

If you want to know how I came to these conclusions, keep reading…

I’m using an one-way analysis of variance (ANOVA) test in IBM SPSS since I want to find out if there’s a statistically significant difference between more than two groups. The year is my independent categorical variable, and the word count is my dependent continuous variable. My null hypothesis is that the means for all four years are equal.

First, I need to check the assumptions. The word count variable is continuous. The pairs (month and word count) are related. The observations are independent (each month’s word count is not influenced by any other month). Check, check, and check. Now, as for normality…

The Kolmogorov-Smirnov statistic is greater than 0.05, indicating the assumption of normality has not been violated.

The histogram confirms the data is distributed in that normal bell-shaped curve. I didn’t really expect this data to be normal, so it just goes to show how often this distribution naturally occurs!

The Normal Q-Q Plot shows the data points are distributed quite close to the trendline. The two outlier months are quite obvious.

On to the analysis!

I can see the average word count for each year is pretty different, as I expected.

Sig. for Levene’s test are all greater than .05, meaning I have not violated the assumption of homogeneity of variance.

In the ANOVA table, the Sig. value is less than my alpha level of 0.05, meaning there is a significant difference in the mean scores on the dependent variables.

Digging into the Multiple Comparisons table, Tukey’s post-hoc test shows that the mean difference between year 1 and 5 is significant at a 0.05 alpha level.

This isn’t a bad result at all. After all, year 1 was my highest word count by far. I don’t mind falling short of that number, even by a statistically significant amount. It’s pretty cool that my word count didn’t drop significantly between year 3, 4, and 5, since I wasn’t in school, took two classes, and then returned to school full-time.

Another interesting thing to note is that when I do this analysis by calendar year, including all twelve months of data instead of just eight, Sig. is 0.460, which is greater than the alpha level of 0.05. That means there’s actually no significant difference in the ANOVA.

Next time, we’ll look into how my word count affects my earnings!