Innovation in estimation
A reliable approach for radio audience indicators
Aurélie Vanheuverzwyn & Joël Chaskalovic
Abstract
Until the end of 2005, the Radio audience indicators' precision was estimated thanks to formulas based on normality hypothesis that is not theoretically available in the case of complex sampling plan. In this context, a research program was launched in 2005 which objective was to produce a reliable estimation of confidence intervals with "easy to use" formulas for each Radio audience indicator.
The alternative method which will be presented at WM3 is composed of two steps. On the one hand, is used the bootstrap technique which consists in drawing a large number of random samples with replacement, called bootstrap samples, from the original sample. Therefore, one obtains an empirical distribution of the indicators of audience, which enable to build confidence intervals for each indicator, station, day part and target.
On the second hand, is built a model for the confidence interval based on a mutliplicative structure composed by the value of the indicator and the number of the individuals which belongs to the target. This first choice of explanatory variables is based on descriptive statistic techniques.
Finally, the method is statistically secured by data mining exploration to investigate other potential dependency of the confidence interval, regarding the other avaliable variables, e.g., the target, the station and the day part. To acheive this objective, we implemented neural netwoks which showed us that the model described above is very stable against the other explicative variables.
Aurélie Vanheuverzwyn
Mediametrie, France
Aurélie Vanheuverzwyn is Scientific Director of Médiamétrie.
She joined Médiamétrie in 1999, where she became Chief of Measurement Science of the Internet Department in 2001 and then Scientific Director in 2003. She contributes regularly to conferences on statistic and survey methodology.
She is graduate of ENSAI, French engineering school of statistic and information sciences.
Joël Chaskalovic
Zenith Optimedia, France
PUBLICIS GROUP Since April 1993
Director of Methodology and Media Research, Publicis Groupe Media France (2003-2005)
Co-Director in ROI Project (Return On Investment), Geo-marketing and Media Database Intelligence.
Director of CRM and New Technologies, ZenithOptimedia France (2001-2002)
Business Development in CRM (multi-channels combination for Business and Media Intelligence: Media efficiency, e-mailing Mining analysis, segmentation, typology, etc).
Director of Data Mining, Publicis Technology (1999-2000)
Conception & Realization of Data Mining models for profiling and extrapolation of marketing target “likely Avantime” (Renault).
Director of Scientific Research, Publicis Group (1993-1998)
Optimization of Media planning Television: Conception of a specific tool for advertising space buying.
R&D team management (6 people in the team).
UNIVERSITY PIERRE AND MARIE CURIE (PARIS VI, JUSSIEU) Since September 1990
Associate Professor
- Mathematical modeling applied to Engineering Sciences, Media and Marketing, Medicine:
- Data Mining applied to Media and Marketing models (Audience forecasting, Distributions of contacts, Law of memorization, etc.)
- ROI modeling, models of gravity applied to Geo-Marketing optimization.
- Medical applications in Urology and Cancer (Quality of Life qualification regarding sexual function following radiotherapy or radical prostatectomy: statistical evaluation by UCLA questionnaire).
- Fluid Mechanics applied to Meteorology.
- PhD direction in modeling and applied mathematics.
- Scientific direction for Bachelor and Master’s degree of Sciences.
|
KEEP ME INFORMED |