Arbitron To Kabrich: 'You're Wrong' -- Kabrich To Arbitron: 'Oh Yea?'
July 2, 2009 at 7:30 AM (PT)
ARBITRON has responded to consultant RANDY KABRICH's charge from MONDAY (NET NEWS 6/29), when he rolled out a follow-up to his chart of a year ago concerning Urban listening tracking in line with the number of Urban Respondents in the Total Sample of HOUSTON's PPM.
KABRICH then revised his numbers YESTERDAY (NET NEWS 7/1) and told ALL ACCESS, "ARBITRON had indicated to several trades that the sample size on the graph released MONDAY was wrong, but did not release any corrected numbers. Upon triple-checking (and finding a discrepancy between two different PPM software vendors), the numbers for the last few months of the station AQH changed ever so slightly. The sample size was always correct. The slightly revised graph show the revised numbers were at the station level, not the sample level -- which does not change the essence of the chart -- in fact, it makes the correlation even more apparent now."
Arbitron has top-notch statisticians on staff, but we donâ??t need that kind of help in this case. The numbers speak for themselves. Trying to force a visual correlation from data that isnâ??t correlated doesnâ??t work.
ARBITRON begs to differ, releasing this statement TODAY:
"We ran the numbers and the answer is: No, the Black 6+ AQH numbers and the size of the Black 6+ sample in HOUSTON during the relevant period are not significantly correlated. The idea of correlation between two or more variables is pretty simple. If they are correlated, when one goes up, the other goes up. When one goes down, the other goes down. That isn’t happening here.
"Over the course of the months in these trends, there are 28 changes in sample size and most of the time, KMJQ’s AQH ratings move in the opposite direction of the sample changes. KBXX’s AQH ratings move in the opposite direction nearly half the time. There are six occurrences where all the variables (sample size and AQH for both stations) move in the same direction.
"ARBITRON has top-notch statisticians on staff, but we don’t need that kind of help in this case. The numbers speak for themselves. Trying to force a visual correlation from data that isn’t correlated doesn’t work.
"And to clear things up, the sample sizes weren’t off, but the station estimates were. At KABRICH’s request, we looked at the comparison of his CORRECTED chart."
ARBITRON's figures can be viewed here.
Not Surprisingly, Kabrich Disagrees With Arb's Disagreement
Said KABRICH, "It appears that ARBITRON should have probably used their top-notch statisticians such as Dr. ED COHEN that they claim to have on staff instead of leaving it to JOHN SNYDER. There are plenty of formulas that prove correlations and disprove correlation. Unfortunately, ARBITRON chose to look at the single month -- not the relationship of multiple months on each other.
"For example, using ARBITRON’s example in their response, if a station trended 2.5 - 2.4 - 2.7 - 2.6 - 2.9 - 2.7 - 3.0 - 2.9 - 3.1 - 3.0 - 3.2 - 3.1 - 3.5 - 3.4 - 3.3 - 3.5 etc , even though the station is up from a 2.5 - 3.5, using the simplistic calculations that ARBITRON provided, they would state the station DID NOT SHOW UPWARD MOMENTUM as the station had as many down months as up months. Clearly, they cannot see the forest for the trees, as no one would argue that the trend line is flat.
"If this is the kind of research we get out of ARBITRON, it appears that the new CEO has some additional house cleaning to do."
Correcting KABRICH, ARBITRON noted, "He's wrong. We looked at every month. And, to his point, Dr. ED wrote that piece."
"Just like PERRY MASON, let me point out in response #1 'ARBITRON has top-notch statisticians on staff, but we don’t need that kind of help in this case.' In response #2 'And to his point, Dr. ED wrote that piece,'" quipped KABRICH. "Can ARBITRON not even keep their story straight for four hours?"