From: "Stephen Wolf" To: "VQEG" Subject: SSCQE vs. DSCQS Date: Thu, 5 Apr 2001 09:00:09 -0600 X-Priority: 3 (Normal) Importance: Normal Dear Colleagues, After examining SSCQE and DSCQS data, I am convinced that SSCQE is a suitable choice for the NR-RR tests provided these tests are designed in a balanced manner that is suitable for statistical analysis (i.e., analysis of variance - ANOVA). With this in mind, I am proposing the following experiment design: 1. Scenes - 6 carefully selected 1 minute segments. 2. HRCs - 10 total (1 original, and 9 processed versions selected from the total set of 40 possible HRCs). This will produce a total of 60 minutes of SSCQE video. To assure that all the viewers see all the video, I am proposing that each subject view this 60 minutes of video using two 30-minute sessions, separated by a break (each tape will also have a short stabilization segment at the beginning). Multiple randomizations are desired so we will need to edit more than 2 viewing tapes. Preferably, 3 sets of tapes should be used (lets call these the red, green, and blue tapes). Subjects should be randomly assigned to one of the 6 possible orderings (R1-R2, R2-R1, G1-G2, G2-G1, B1-B2, B2-B1). Each lab should have an equal number of subjects at each ordering. Perhaps 3 subjects per ordering, for a total of 18 viewers per lab. This test design is about the minimum that I believe is acceptable to form the basis for an international recommendation. Stephen Wolf ================================================================ From: "Stephen Wolf" To: "VQEG" Subject: RE: SSCQE vs. DSCQS Date: Thu, 5 Apr 2001 11:32:14 -0600 X-Priority: 3 (Normal) Importance: Normal Beau, I would not want the session time to increase. I would concatenate all the scenes together and discard the first 8-10 seconds of each scene in the data analysis (to allow the viewer's SSCQE to stabilize to the new scene x HRC combination). Of course, this would create 50 second scenes in the end rather than 60 second scenes, but I think we could live with that. With this view in mind, the tape randomization should be done considering each 1 minute scene x HRC combination a unique sample. S.W. > -----Original Message----- > From: Andrew Watson [mailto:abwatson@mail.arc.nasa.gov] > Sent: Thursday, April 05, 2001 10:23 AM > To: Stephen Wolf; VQEG > Subject: Re: SSCQE vs. DSCQS > > > Stephen, > > I assume you would want gaps between 1 minute segments, and rest > breaks within the 30 minute session (5 minutes every 10 sequences?), > so that the actual session time might be closer to 50 minutes, or a > total of about 2 hours/subject? > > -beau