This extension enhances Google Analytics by providing implicit statistical evaluation of data.
One very useful feature of Google Analytics is the comparison of timeframes for metrics. You can compare number of sessions, bounce rate, E-commerce transactions, AdWords cost, etc. between adjacent weeks, months, or years. One shortcoming in this comparison feature is that the percentage change calculated by Google for these metrics is not qualified. This extension addresses this shortcoming by evaluating the change between time periods for statistical significance. Currently, for 5 < points < 40 the Wilcoxon Signed-Rank test is used. For points > 40, the paired Student's t-test is used. Release Notes: Version 3.1 * Separate tab for each metric; support multiple metrics on graph * Should support metrics from line graphs on any page * Reload button for when extension misses capturing data * Also show date range and percentage change for metric Version 2.3 * Update calculation of Wilcoxon statistic (used when n<40) with port of SciPy implementation * Styling changes * Warnings added and other fixes Version 2.1 * Versions 1.* were broken due to changes in Google Analytics. Updated data extraction to pull directly from Google's web requests, rather than calculating from the DOM. More accurate and robust * When you change the dates or metric on a chart, it will automatically recalculate Limitations: 1. Time frames to be compared should begin on the same weekday (Monday and Monday or Friday and Friday), to minimize confounding factors in the comparison. They should be the same length of time. Be careful to ensure this, especially for year-over-year monthly comparisons. This script derives its statistical power by treating metrics as paired data points (based on the day of the week). 2. Must compare at least 5 data points (days, weeks, or months). Beyond the lower limit, the statistical test lacks sufficient power. For further documentation and description of the statistical tests, visit https://noahonnumbers.com/blog/entry/statistics-google-analytics
- (2020-04-02) Eric Richard: The extension only says that "data is loading". Seems not to work. Sorry for the rating but chrom web store wants us to rate first before we write the review.
- (2014-05-30) Sean Clark: I am used to using "statistical significance" in our conversion rate optimisation, but to have it easily available directly in Google Analytics for any data is just awesome.
- (2015-03-20, v:1.3.2) Paul Burdon: N/a
I have the same n/a problem. Any ideas how this could be resolved?
- (2014-12-18, v:1.3.2) james zanza: N/A
Hi Noah, I've implemented the tool but it shows me always N/A for every comparison I try...what am I missing?? Mant thanks, James
- (2014-06-10, v:1.3.2) Christopher Mason: Time Range not being evaluated correctly
Hi Noah, I recently added your Chrome extension, however, I've found an issue. I've been struggling to get the tool to find any significance, so did a test of a data set I know to be significantly different - in this case 'Transactions' over two 4-week periods, comparing Mon May 5 - Sun Jun 1 with Mon Apr 7 - Sun May 4 (ie. the previous 4 weeks). The tool threw up the following warning & significance estimation: ___________ GOOGLE ANALYTICS STATISTICAL SIGNIFICANCE Warning: time periods begin on different days. Match starting weekdays for valid comparison. Test type Wilcoxon Signed Rank Metric compared Transactions Number of points 28 P-value > 0.10 Significant difference? NO ____________ Could it be a time-zone issue? I'm based in Melbourne, Australia, and we often see issues related to this. The raw data is below - as you'll see, these data sets evaluate as being significantly different using the Wilcoxon Signed-Rank test. Cheers, Christopher ______________ 176 406 180 407 157 500 148 598 187 490 86 278 76 307 215 647 181 607 157 707 163 652 83 790 65 483 74 629 122 1754 259 2116 272 2283 223 576 124 142 93 64 114 52 325 202 314 173 298 174 318 101 297 105 159 36 195 65