Skip to main content
21 events
when toggle format what by license comment
S Nov 17, 2014 at 17:55 history bounty ended Livid
S Nov 17, 2014 at 17:55 history notice removed Livid
Nov 17, 2014 at 14:18 vote accept Livid
Nov 10, 2014 at 23:39 history tweeted twitter.com/#!/StackStats/status/531954427746201602
Nov 10, 2014 at 19:28 answer added Tom Minka timeline score: 3
S Nov 10, 2014 at 18:11 history bounty started Livid
S Nov 10, 2014 at 18:11 history notice added Livid Draw attention
Nov 10, 2014 at 17:16 history edited Livid CC BY-SA 3.0
fixed typos
Nov 8, 2014 at 3:25 comment added Livid @whuber Sorry, I cannot risk revealing the data/purpose at this point. As I said, use the free throw scenario. I found it very difficult to find examples of processes that lead to underdispersion regarding any subject matter, so would welcome any examples. So, please do three models. Even if they are not relevant to my description of the system on first glance I am still interested.
Nov 8, 2014 at 3:10 comment added whuber Any "process that can result in this type of data" which incorporates random variables is a probability model. I can hypothesize loads of such models, but without specific information concerning the actual subject matter any of these would be pure speculation.
Nov 8, 2014 at 2:39 comment added Livid @whuber Ah, no that is not the problem here. I used that phrase to mean "the maximum performance achievable by the individual".
Nov 8, 2014 at 2:05 comment added whuber If the data reflect maxima of sets of values, then they will tend to be narrowly dispersed and a little negatively skewed. A binomial model will not describe them.
Nov 8, 2014 at 0:26 comment added Livid @whuber On rereading I realized I do not follow you here. Can you expand on this statement "Max performance level achieved"... suggests that data might not appropriately modeled as an iid Binomial sample.
Nov 7, 2014 at 20:50 history edited Livid CC BY-SA 3.0
response to comment
Nov 7, 2014 at 20:18 comment added gung - Reinstate Monica We may not be able to identify the mechanism that produces your results. The autocorrelation idea is interesting. To investigate it you will need to get the trial by trial data from your study, or run a new study to get such data.
Nov 7, 2014 at 20:08 history edited Livid CC BY-SA 3.0
clarified in response to comments
Nov 7, 2014 at 19:33 comment added whuber Thanks for the update. "Max performance level achieved" may be a key phrase, because it suggests that data might not appropriately modeled as an iid Binomial sample. Unfortunately, your question about "mechanisms" cannot be answered until you disclose what the data really represent! The factors that might affect consistency of free throw performance will be largely different than those that affect consistency of ranks in a series of races, for instance. Having said that, there are plenty of ways to construct probability models of underdispersed phenomena. Maybe that's what you are looking for?
Nov 7, 2014 at 19:28 history edited Livid CC BY-SA 3.0
completed question
Nov 7, 2014 at 18:58 comment added whuber The only question that appears in all this is #1, whose answer you obviously already know. What do you really want to ask? Whether this dataset looks significantly underdispersed (as suggested by the title)? What does the title mean by "mechanisms"?
Nov 7, 2014 at 18:58 history edited gung - Reinstate Monica CC BY-SA 3.0
formatted; light editing
Nov 7, 2014 at 18:50 history asked Livid CC BY-SA 3.0