initial analysis

There is no significant difference between the experimental treatment groups (anova, using R's aov method) when looking at percentage change in weekly use by each user.

> aov1 <- aov(percentage~week_number*experimental_group, data=m[m$se_class=='low', names(m)])
> summary(aov1)
Df Sum Sq Mean Sq F value Pr(>F)
week_number 1 9185 9184.9 12.8365 0.0003446 ***
experimental_group 2 141 70.7 0.0988 0.9059019
week_number:experimental_group 2 1376 688.2 0.9617 0.3823317
Residuals 3443 2463570 715.5
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
> aov1 <- aov(percentage~week_number*experimental_group, data=m[m$se_class=='high', names(m)])
> summary(aov1)
Df Sum Sq Mean Sq F value Pr(>F)
week_number 1 12517 12516.9 17.8116 2.499e-05 ***
experimental_group 2 1199 599.5 0.8530 0.4262
week_number:experimental_group 2 379 189.5 0.2696 0.7637
Residuals 3592 2524243 702.7
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1