The blog continues to grow in site views. I had a little north of 90,000 site views over the entire year. (If you find that impressive don’t be, a very large proportion are likely bots.)

The trend on the original count scale looks linear, but on the log scale the variance is much nicer. So I’m not sure what the best forecast would be.

I thought the demise had already started earlier in the year, as I actually saw the first year-over-year decreases in June and July. But the views recovered in the following months.

So based on that the slow down in growth I think is a better bet than the linear projection.

For those interested in extending their reach, you should not only consider social media and creating a website/blog, but also writing up your work for a more general newspaper. I wrote an article for The Conversation about some of my work on officer involved shootings in Dallas, and that accumulated nearly 7,000 views within a week of it being published.

Engagement in a greater audience is very bursty. Looking at my statistics for particular articles, it doesn’t make much sense to report average views per day. I tend to get a ton of views on the first few days, and then basically nothing after that. So if I do the top posts by average views per day it is dominated by my more recent posts.

This is partly due to shares on Twitter, which drive short term views, but do not impact longer term views as far as I can tell. That is a popular post on Twitter does not appear to predict consistent views being referred via Google searches. In the past year I get a ratio of about 50~1 referrals from Google vs Twitter, and I did not have any posts that had a consistent number of views (most settle in at under 3 views per day after the initial wave). So basically all of my most viewed posts are the same as prior years.

Since I joined Twitter this year, I actually have made fewer blog posts. Not including this post, I’ve made 29 posts in 2018.

2011  5
2012 30
2013 40
2014 45
2015 50
2016 40
2017 35
2018 29

Some examples of substitution are tweets when a paper is published. I typically do a short write up when I post a working paper — there is not much point of doing another one when it is published online. (To date I have not had a working paper greatly change from the published version in content.) I generally just like sharing nice graphs I am working on. Here is an example of citations over time I just quickly published to Twitter, which was simpler than doing a whole blog post.

https://twitter.com/CrimAndyW/status/998566125381324801

Since it is difficult to determine how much engagement I will get for any particular post, it is important to just keep plugging away. Twitter can help a particular post take off (see these examples I wrote about for the Cross Validated Blog), but any one tweet or blog post is more likely to be a dud than anything.

So the blog has continued to show linear growth in terms of views over time, I take a good hit though in December.

I only ended up writing 35 new posts in 2017 (that includes things that are not blog posts, like pages I created for new classes). For comparison in 2015 I wrote 50, and 2016 I wrote 40. I’ve managed to be pretty consistent though over time, here is the cumulative total over time.

That is more or less what I aim for, to just have some content every few weeks.

There is not much to say in terms of popular posts on the site for the year. My most popular posts are ones I’ve written in previous years. I did not have any post this year gain a large number of viewers when it was first written. It is just a slow accumulation of around 200 views per day, mostly people being referred via Google searches.

I wanted to analyze the topics I’ve written about over time, so I grabbed all of the tags I’ve placed on posts. I collapsed both categories and tags, as I don’t really make much of a distinction when I pick them. Here is a graph of the number of posts that have that tag and the page views (this will double count page views, for example a post could have both SPSS and Data Visualization). None refers to pages that are not blog posts, like my home page and pages I created for class syllabi.

If we look at the ratio though, you can see my scholarly posts are mostly ignored, only in total do they accumulate much viewing.

My posts on showing how to use various google maps services with python must be reasonably high in Google searches, as I get a slow trickle of hits for them every day. The high uncertainty is driven by my ratios need to be plotted on log scales post.

I tried to analyze whether the content of my posts substantively has changed over time. I suspected since I took my job at Dallas my posts swayed more towards paper/scholarly (the tag I use for academic related things) and more away from technical computing stuff. I have too few of posts though (and too many tags) to easily make sense of it. Taking only tags that are included on 30 or more posts, here are the counts of those tags over time.

About the only clear trend is that scholarly has risen with SPSS dropping, the other frequent categories though look to me to be fairly consistent. I could spend more time grouping the tags into thematic content, but I have too many other things I need to do (including writing other blog posts)!

Happy New Year!