Statistics Help @ Talk Stats Forum - Applied Statistics
http://www.talkstats.com/
Statistics in finance, economics, engineering. Actuarial science. Econometrics. Operations research.enFri, 25 Jul 2014 14:48:24 GMTvBulletin60http://www.talkstats.com/images/misc/rss.pngStatistics Help @ Talk Stats Forum - Applied Statistics
http://www.talkstats.com/
What is and is not a convenience sample
http://www.talkstats.com/showthread.php/56923-What-is-and-is-not-a-convenience-sample?goto=newpost
Thu, 24 Jul 2014 17:45:07 GMTIt is well accepted that 'convenience samples" can not be used to make statistical inference, essentially because they are not a random sample of the population [and thus can not be assumed to resemble that population].

But what if your entire population is invited to respond to the internet, for a survey? Response rate may be low, as it can be in a random sample, but at what point does this type of survey become a convenience as compared to probabilistic sample?

Is the key how many actually recieve the invitation, how many respond, how many have access to the internet or what?
]]>Applied Statisticsnoetsihttp://www.talkstats.com/showthread.php/56923-What-is-and-is-not-a-convenience-sampleAverage in semi dependent observations.
http://www.talkstats.com/showthread.php/56917-Average-in-semi-dependent-observations.?goto=newpost
Thu, 24 Jul 2014 14:12:10 GMTI have a dataset of measurements (i.e., differences; which are continuous), with some individuals contributing a single measurement while others have...I have a dataset of measurements (i.e., differences; which are continuous), with some individuals contributing a single measurement while others have multiple measurements.

These data could be called non-subsequent collected measures, some are back to back measurements while others are not. I want to control for subject random effects in the calculation of the mean of the variable. What is the best way to report the mean of these measurements while also taking into account perhaps clusters. I want to keep all of the values in the dataset and just weighting them in a mean calculation does not quite seem right.

Can I make a multilevel model and kick this value for the mean out along with confidence interval?
]]>Applied Statisticshlsmithhttp://www.talkstats.com/showthread.php/56917-Average-in-semi-dependent-observations.How to assess impact over time
http://www.talkstats.com/showthread.php/56902-How-to-assess-impact-over-time?goto=newpost
Wed, 23 Jul 2014 22:04:51 GMTI am looking at rehab rates (which are coded either successful or unsucessful)) and factors which influence them. This is a case where both the rehab...I am looking at rehab rates (which are coded either successful or unsucessful)) and factors which influence them. This is a case where both the rehab rate (the response variable) and the predictor variables are taking place over time. So I don't think I can just use logistic regression, since time series is involved I will get spurious regression if I do. I assume some time series method has to be used.

There are clear trends in the rehab rate over time and (likely) the predictors as well.
]]>Applied Statisticsnoetsihttp://www.talkstats.com/showthread.php/56902-How-to-assess-impact-over-time