tag:blogger.com,1999:blog-5847937807428508775.post1423169432942509406..comments2013-08-26T03:06:24.679-07:00Comments on Mokken scaling and item response theory: R syntax for Mokken Scaling AnalysisRoger Watsonhttp://www.blogger.com/profile/07408659363029289322noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-5847937807428508775.post-21028969812485364902013-08-26T03:06:24.679-07:002013-08-26T03:06:24.679-07:00Sorry I did not notice this question; I am also so...Sorry I did not notice this question; I am also sorry I can't really help as I am more or less totally dependent on Andries van der Ark for the syntax and any other advice - without him. R is just a mystery to me! He's very helpful if you email him.Roger Watsonhttps://www.blogger.com/profile/07408659363029289322noreply@blogger.comtag:blogger.com,1999:blog-5847937807428508775.post-52814941671129360172013-05-28T14:50:52.696-07:002013-05-28T14:50:52.696-07:00Thanks for your reply! I will continue to think ab...Thanks for your reply! I will continue to think about the problem.<br /><br />In the meantime, I do have another question about the plotting function for iio in the R package (if you don't mind):<br /><br /> I am having trouble figuring out how the function determines the order in which the plots are presented. If I use (item.pairs = x), it seems logical for the first few items, but then the ordering seems odd. Any thoughts on this?stefaniehttps://www.blogger.com/profile/10140377396824874628noreply@blogger.comtag:blogger.com,1999:blog-5847937807428508775.post-68012832626960823332013-05-28T08:56:02.120-07:002013-05-28T08:56:02.120-07:00Dear Stefanie
Very glad you found the blog useful...Dear Stefanie<br /><br />Very glad you found the blog useful. I am afraid that there is not way within the Mokken scaling packages that I know of for taking into account missing data. It likes a complete data set so I usually just delete listwise the ones that are incomplete. Of course, if the missing data are random across the data set you can lose a lot of people. The only way to save the data that I know of is to impute - SPSS has a few very complicated options that I don't understand! The other way is to insert either a neutral value, if your data are polytomous, or to insert the mean value for the group for missing variables. However, these strategies depend on him mush is missing, as an estimate I'd say don't do this for missing data more than ten percent. I'll be very interested to hear more about your work. I am just reviewing a manuscript from a team in USA using Mokken scaling, unusual for the USA, it's usually only used in Europe.<br /><br />RogerRoger Watsonhttps://www.blogger.com/profile/07408659363029289322noreply@blogger.comtag:blogger.com,1999:blog-5847937807428508775.post-37158500623437628432013-05-28T07:09:39.471-07:002013-05-28T07:09:39.471-07:00Hi Roger,
Thanks for this post, and for your blog...Hi Roger,<br /><br />Thanks for this post, and for your blog! I am using Mokken scaling and the mokken R package for my dissertation research.<br /><br />I am encountering some trouble with Mokken analyses of a set of rating scale data because of the fact that not every rater rated each item. Have you encountered an incomplete data set in a Mokken analysis? Any thoughts about what to do?stefaniehttps://www.blogger.com/profile/10140377396824874628noreply@blogger.com