|Saturday, February 22|
|CS22 Survey Analysis||
Sat, Feb 22, 10:45 AM - 12:15 PM
Forecasting Panel Turnover Utilizing Survival Analysis (302761)*David Burtnick, Nielsen Audio
Keywords: panel, forecast, data censoring, survival analysis
Accurately forecasting turnover for a survey panel is the key to maintaining panel size over time, allowing for appropriate budgeting and allocation of resources. We have been able to consistently forecast turnover in Arbitron’s PPM panel with an error less than 1%/mo. A panel can be partitioned based on criteria that influence tenure such as whether an incentive is given. A survival analysis can be performed on each segment. Appropriately censoring certain data is crucial to the analysis. The results of the survival analysis can be used to define a probability function of the likelihood of a panelist remaining in the panel for a specific time period. The time period can be a day, week, or month, depending on the amount of data and size of the panel. The product of the probability function for consecutive time periods is the probability of remaining in the panel for the entire time frame encompassed by the time periods. Subtracting this result from one gives the probability of leaving the panel during the time period. This probability, applied as a weight to all of the panelists in a segment, yields a turnover forecast. This approach is illustrated using empirical data.