I implemented a function to compute p-values from z-scores, so now it will tell you whether the result is statistically significant or not.
Again, the interface could surely be improved, but the values are now calculated.
I implemented a function to compute p-values from z-scores, so now it will tell you whether the result is statistically significant or not.
Again, the interface could surely be improved, but the values are now calculated.
Tal Galili 1:30 am on July 16, 2010 Permalink |
Good job.
idea 1: I would suggest to use colors to indicate when it is and isn’t significant (red for not, green for yes significant) – only on the text that says “are/not confident”.
idea 2: I do wonder how how many digits after the point are relevant. I believe that beyond the standard deviation, it is not that relevant.
Keep going, and thank you 🙂
Tal
Mark 1:31 pm on July 18, 2010 Permalink |
I have a red-green color deficiency, like 7-10% of males. Please do not rely only on color patches to distinguish a status or similar.
mitcho/芳貴 11:16 pm on July 18, 2010 Permalink |
Colors for significance is good, but I’m also hoping to add clearer information on the percentage difference in the metric, as compared to the control.
Mark, I’ll definitely make sure that colors are only an aide and are not the only way information is conveyed.
Using red and green is also dangerous because they have different cultural connotations as well. In most of the west and in Japan, red = negative, green = positive, but in China, red = positive, green = negative:
Note the left column. So stocks which are going down are in green. (My apologies, Mark, this image may not be quite as interesting to you.)