Nielsen confirmed as much when some clients asked it to explain the phenomenon of declining demographics that has been occurring when it adds DVR playback data to its so-called "live" ratings. The live ratings, of course, were never entirely "live" and have included estimates for VCR audiences for about two decades. It is how Nielsen imputes those VCR audiences that have caused the anomalies, the company said.

"With time shifted data, there have been occurrences when demographic projections (in units) for 'live plus same day' are lower than for 'live' projections (units)," Nielsen said in an explanation issued to some clients after they complained about the phenomenon. "This difference can be attributed to the imputed VCR record activity which is calculated from household tuning activity that includes VCR record and household tuning that excludes VCR record. The VCR adjustment factor is applied to each building block demographic at the quarter-hour level for both programs and time periods."

In simple terms, Nielsen spokesperson Anne Elliot said the math created by Nielsen to impute VCR audiences was designed for its "live" only ratings stream and when applied to the relatively small samples in its new DVR data streams "changes the numerator and denominator," yielding aberrations like those discovered by some clients.

In its explanation, Nielsen provided an example using the Dec. 29, 2005 episode of CBS' "Criminal Minds," which because of the math actually generated a negative DVR ratings factor, causing the show to lose seven viewers in the women 21-24 demo after the adjustment.

Nielsen's Elliot said the ratings company is still trying to understand all the phenomena that are occurring as it begins processing the multiple ratings streams, but that many of those aberrations should diminish as the DVR sample grows and becomes more representative of the total population.

. "Obviously, there is a logic test this does not pass," she conceded, adding, "It is clearly also related to the fact that the sample is so small."

At least one knowledgeable observer, Lifetime Television Executive Vice President-Research Tim Brooks, believes similar anomalies may be occurring because of the weighting schemes Nielsen uses to adjust its ratings based on the composition of certain demographic groups in its samples. If important demos are underrepresented in its sample, Nielsen frequently uses mathematical factors, giving them more "weight," and raising audience levels ascribed to those demos in its final ratings.

While that practice is accepted by most Nielsen clients, Lifetime's Brooks says it could be contributing to some of the distortions as they are applied across multiple ratings streams.

"Nielsen's calculation systems are so complex, especially since they've weighted people in the database so that a person is not a simple person. A person is 1.1 on one day and 1.5 on another day," Brooks explains, adding, "Therefore, when they do multiple calculations they run the risk that people who were heavily weighted in the 'live' sample, may not be in the playback sample."

In fact, a MediaDailyNews analysis of data released to the press for the first week of DVR-included ratings, revealed some interesting vagaries. For example, figures for two programs--"Two and a Half Men" and "King of Queens"--show the seemingly impossible: more viewers on the day the show aired than in the "live plus seven day" tally.

Problems also existed in the DVR ratings for total viewers. Again, the seemingly impossible was the case with fewer "Two and a Half Men" viewers watching the show in the "live plus seven day" category than in "live plus same day." The same occurred for episodes of "Criminal Minds," "Cold Case," "60 Minutes" and others. Also, two CBS broadcasts appeared in the "live plus same day" ratings and not at all in the "live plus seven day:" an "Evening News" special and the "Sun Bowl." Notably, the problems appeared to be all with CBS shows.

This news story is important. I recommend that everyone read it.