Add error bars (using /- 2 standard error) to the herceptin and perjeta bars. does this change your interpretation from question 4? are herceptin and perjeta statistically significantly different?