Chrome has recently overtaken Firefox as the second most popular Internet Browser (after Internet Explorer). Many users find it faster and more convenient. In addition, supporting Firefox on servers like Winstat has become very difficult because it is updated so frequently.
We've thus installed Chrome on Winstat, and we plan to remove Firefox during our next major Winstat upgrade. This should take place during the upcoming semester break. Most SSCC staff members have already switched to Chrome on their own computers and we think you'll be happy with it.
If you have bookmarks in Firefox, Importing Bookmarks/Favorites into Chrome will help you move them to Chrome. Please do so at your earliest convenience. If you use Firefox add-ons, first check to see if there's a corresponding Chrome add-on. If not, please contact the Help Desk and we'll see what can be done.
SSCC staff are working with Microsoft to figure out why individual Winstat servers occasionally stop accepting logins. We've developed an automatic procedure that makes sure users are not directed to problem servers when they first log in. However, if you have a disconnected session on a server that won't accept logins, you won't be able to reconnect to it, and you won't be able to create a new session until the existing session is reset automatically.
We've thus temporarily reduced the amount of time before a disconnected Winstat session is reset from three hours to one hour. This will reduce both the probability of the problem occurring and the duration of the problem if it does occur. Once the problem is resolved, we'll increase the time back to three hours again.
If you are running long jobs and believe this change will cause problems for you, recall that the SSCC has a special Winstat server specifically for long jobs. It never disconnects idle sessions or resets disconnected sessions. If you'd like to use it, contact the Help Desk.
Stata 12 introduced the mi impute chained command, which allows you to impute data using the ICE method. mi impute chained can do essentially everything Patrick Royston's ice command can do, and it's easier to use. It also allows you to use the entire universe of mi commands. mi estimate is the replacement for mim, but mi also has commands for postestimation, changing multiply imputed data, merging multiply imputed data sets, and much more. The mi data format is also more efficient and flexible than that used by ice/mim.
We thus recommend that you switch from ice to mi, while expressing our gratitude to Patrick Royston for making ICE available in Stata for so many years. To help you make the transition we've created a new Knowledge Base article Multiple Imputation using ICE in Stata.
We are concerned, however, that some SSCC members are using multiple imputation without spending enough time learning the theory behind it. Multiple imputation using ICE is somewhat controversial as it lacks a formal theoretical justification. The results can easily be invalid if the imputation model is not chosen properly. We strongly recommend a thorough reading of the literature before using multiple imputation.