Closed AmberSJones closed 9 years ago
Can you give me an example of this?
We forgot to discuss this today (1/22). I think we'd need to get with Doug or Courtney to see an example calculation. At this point, the sample sizes are not large enough for statistical significance.
Update 4/8: When Doug sends the new data file, he'll provide an example of a determination of statistical significance. We will then need to implement on-the-fly calculation for cross-dimensional questions (i.e., a 5x5 matrix of responses). In these cases, we'll implement the flag as a positive or negative warning sign in the bottom left corner (intersection of x and y axes).
In addition to this feature, the heatmap now uses a cutoff of 5 respondents as requested in our last meeting.
The flag has been implemented.
Ok- I want to do a few checks on this... there are a few cases where the result was not significant that I was curious about.
Also, I wonder if we should put the text next to the flag icon to make it more obvious since the hovering is slow.
Ok, I did some spot calculations, which verify the result that you're showing.
What do you think about putting text next to the flag? We might need input from @horsburgh and Doug.
So - there is a single flag that tells me whether the whole matrix is significant? I'm not sure how to interpret the flag. If it is a single flag for the whole table, I think you should just put the text next to the flag in the same box. Plus, it would be good if the text was a link to a pop up or something that has a paragraph describing the test that is used to determine whether it is significant or not.
Yes, a single flag for the whole matrix to say whether the results of that matrix are significant or not. I think both of your suggestions are good. Maybe Doug can compose the text- or I could try to extract it from his email.
The flag now has the text next to it instead of on hover. I changed the style a little as well. Let me know what text should be displayed on click.
How's this for some text: 'This flag indicates whether the differences in the results in this matrix are statistically significant according to a chi-square test. The distribution of observed frequencies is compared to expected frequencies from a distribution where there is no relationship between the two axes. If the total difference is greater than the chi-square statistic for the 95% confidence level, than the results are considered statistically significant.'
The text has been added and the pop up is triggered when clicking on the flag text.
Works for me. Closing.
We need to implement some sort of flag to indicate whether values are significant or not. This would be calculated on the fly using a fairly simple chi-square statistic.