On my last blog post I looked at which day of the week is better to buy VTSAX over time. On that post I showed that over the time period ranging from 11/13/2000 thru 10/5/2018, the median price difference between the daily price and the weekly average was the lowest on Monday. The plot showing this result is once again shared below

**Box Plot for daily price difference from weekly average:**

Since this post, I have received a few comments asking if the difference is “really real” meaning is this difference statistically significant? One way to answer this question is through hypothesis testing. For those asking a similar question, I decided to share my findings here: Looking at the price variation from the weekly average, the Kruskal-Wallis test rejects the null hypothesis that all days come from the same distribution at a 1% significance level. The Wilcoxon rank sum test also rejects the null hypothesis of equal medians at the 1% significance level between Monday and any other day except for Tuesday. So basically, I would say yes, this difference is statistically significant. Now, while these results help answer this question, I am more of a graphical person. Therefore, I decided to plot the 95% confidence intervals for the median daily price deviation from weekly average. This plot is shown below with blue circles indicating the Median Value and Red Circles indicading the +/- 95% Confidence Intervals (CI). Once again, since the Monday CI does not overlap with the CI of any of the other days except Tuesday, I would say that this difference is significant. Even with Tuesday, the tails of the CI are just barely overlaping.

**+/- 95% Confidence Intervals for Median Daily Price Difference:**

By the way, I have decided the use the median and not the mean here since there are a number of outliers in the data, as seen on the box plot above. The median value would be less affected by these points.