Frequently Asked Questions
What is it?
It's a trend analyzer a bit like Google trends but instead of using search term data (the things people type into the search engine), it uses a search engine to analyse the content of web pages, blogs, etc. This means that we can analyze what people know (and write about) rather than what people want to know (and therefore search for). It's also a lot more fun!.
How does it work?
We find out the total number of web pages (using Yahoo Web Search web services) containing the term "Tuesday" for example. Let's call that A. Then we find out how many pages have both "Tuesday" and "hungry", and we call that B. The fraction B/A is then plotted on the graph above Tuesday. Repeat this for the other days of the week, and also for "thirsty" and you get this result. Actually it's a bit more complicated than this, but this is the general idea.
What can I do with it?
The only limit is your imagination...
- Which day of the week do people like to clean their fishtanks?
- Has red wine got more popular than white wine in recent years?
- Try some of the random examples to get an idea.
- Read the blog posts for inspiration.
Help, I can't think of two things to type in the boxes...
Don't panic - just leave one box empty.
How do I interpret the results?
With care! Remember, it's all based on the number of web pages containing individual words or phrases. Words in isolation may not have the meaning you expected. For example, while you might think the word "Windows" means the operating system, someone else might think it's something you gaze out of for inspiration. Of course, it's both. There's no perfect technology to allow computers to "understand" text written by humans, but you can get around some of these problems by rewording your search terms (for example, use "Windows XP" instead of "Windows").
How do I blog my cool new plots?
When you create a new plot, two handy links will appear underneath (after "share your plot"). You can use the URL of either link in a blog entry, web page or email. You might want to use a service like tinyurl.com to make the long URLs easier to handle. If you want to embed the image in a blog or message board you may find that the long image URL is not allowed, or that the image itself is too big. You may find the following workaround useful (it's easier than it looks):
- Click on the "image link" (ideally so it opens in a new window or tab) and you'll see the image (now with legend and site attribution)
- Edit the w=400&h=200 part of the image URL to the width and height you need (e.g. w=600&h=300)
- Convert the link at tinyurl.com (or similar site)
- Append #image.png to the tinyurl and use this as the image URL
- Ideally link the image to the original plot page (using the "page link"), as in the example below.
Example:
The following image has src="http://tinyurl.com/2cu2ew#image.png"
On many bulletin boards you would use the following "BBcode": [img]http://tinyurl.com/2cu2ew#image.png[/img]
Does the y-axis start at zero?
Yes, if you check the box under the drop-down menu. Otherwise, the pink and purple plots are rescaled separately to show the maximum amout of change in the available space (note that this can often exaggerate the importance of a trend).
How do I compare trends for other things (on the x-axis)?
We'll happily add suitable new x-axis categories if you get in touch with us (see below). But, you probably didn't know that you can compare anything you like in an almost identical way using the legacy version - but note that the x- and y-axes are switched around. Here's an example for times of day. You're limited to four things with the legacy version and the results may not be quite as "smooth" as the new version, but you've got much more freedom...
More questions and comments?
There are two ways to send your feedback:
- Type a short comment into the text box under the plots on the main page
- Post a message on blog's guestbook - if you don't want your message posted in public, please say so (all first messages go into a moderation queue).
© page contents and layout are copyright 2007 compare-stuff.com