Skip to main content
12 events
when toggle format what by license comment
S Dec 6, 2014 at 1:03 history bounty ended T.J. Gaffney
S Dec 6, 2014 at 1:03 history notice removed T.J. Gaffney
Dec 5, 2014 at 17:29 vote accept T.J. Gaffney
Dec 5, 2014 at 7:40 answer added jvbraun timeline score: 2
Dec 5, 2014 at 0:41 comment added T.J. Gaffney @cdeterman Thanks for your response. In the group I work with we might look at something like age. (Say we're trying to predict likelihood they watch television.) If you want to make a good predictive model, you probably shouldn't consider each year or each day as its own category. But say one person in my group wants to consider 5 year bands, but another wants to consider 10 year bands, and another 15 year bands. Maybe the Chi-Squared test says that age is a predictor of television-viewing, but only with the 10 year bands, should we still use it? What if we take any n year bands?
Dec 4, 2014 at 20:38 comment added cdeterman Generally the only reason to combine classes is because your expected counts are very low (>5). Do you have low data or are you just trying to 'find significance'? Why are you combining groups?
Nov 29, 2014 at 4:29 history edited T.J. Gaffney CC BY-SA 3.0
Changed "qualitative" to "quantitative"
Nov 29, 2014 at 3:09 history tweeted twitter.com/#!/StackStats/status/538530171159343104
S Nov 29, 2014 at 0:28 history bounty started T.J. Gaffney
S Nov 29, 2014 at 0:28 history notice added T.J. Gaffney Draw attention
Nov 26, 2014 at 23:41 review First posts
Nov 26, 2014 at 23:48
Nov 26, 2014 at 23:36 history asked T.J. Gaffney CC BY-SA 3.0