Publication Type: Conference Paper/Unpublished Manuscript
Abstract: Researchers and statisticians working in state agencies are often called upon by governors and state legislators to provide analyses of data to support or refute a particular political initiative. This presentation will explore the analytical and ethical implications of providing analyses that may be used for partisan purposes. Examples from the author’s own experience of 10 years as a state researcher, as well as the experiences of others, will be provided.

Publication Type: Paper/Poster Proposal
Abstract: Arbitron has recently developed a new multi-media measurement method – called the Personal People Meter (PPM). Television ratings obtained using the PPM have been considerably higher than those measured by Nielsen Media Research’s (NMR) set-top diary. To further explore this ratings difference, Arbitron and Nielsen Media Research (NMR) placed stationary PPMs in 19 households that were already outfitted with TV set-top meters in order to monitor how often these PPM’s picked up exposure from TV’s in the same room versus a different room. Although it is currently unknown whether people using the diary/set-top system credit such instances as “watching or listening”, an audio-based measurement system would likely credit such instances. In order to know whether the higher PPM TV ratings are due to a difference in the definition of audience between the two measurement systems we must determine how much the PPM credits out-of-room TV exposure.

Since we did not know where the people were located when the TV’s were tuned, we cross-tabulated separate data sources to derive estimates of how likely people were in the same or different rooms as the TV’s. The data sources consisted of: a survey conducted with the household to determine what times they were usually away from home (e.g. at work or at school) and the NMR people meter data. Using this data, we assigned to each person in the household, their probability of being in each room of the household. These probabilities were overlaid onto the media detections of the stationary PPM’s in each room of the household to determine that if the person were wearing a meter, how many minutes of detected TV exposure would have occurred while the person was in the same room as a TV and how many would have occurred while in a different room than the TV.

Publication Type: Paper/Poster Proposal
Abstract: The National Household Enumeration Survey, an in-person interview conducted in English or Spanish annually since 1994, has been evaluating the idea of converting its current paper-based instrument to computer driven systems. A primary concern in considering conversion to the technology is the potential effect of the collection device on data quality. This presentation will focus on procedures and software features that were developed and evaluated for their impact on data quality. There were concerns that the accuracy of the recorded responses may decrease given the fact that, with the PAPI questionnaire, data recorded on the paper form were key-entered by two professional data entry operators while running a software program that would detect any keying errors. Using the handheld computer, the function of the two keyers would be replaced with a single field interviewer entering the data once with no obvious check for keying errors. Additionally, there were questions regarding the possible impact on response rates and/or the data being provided by a respondent given the presence of the handheld device in place of the paper form. Would the responses differ or be skewed relative to the data collected previously via PAPI? The objective of the evaluations was to identify and quantify the level of impact the change in data collection media had on data quality; specifically regarding completeness and accuracy. Data will be presented comparing responses collected with the handheld computer with that of the PAPI approach. Implications and recommendations for converting from PAPI to electronic field data collection will be discussed.

Publication Type: Conference Paper/Unpublished Manuscript
Review Method: Peer Reviewed
Abstract: The pin-point accuracy of the client computers’ or server computers’ data collection software created a high expectation of the Internet use measurement. However, the relatively unexplored data structures and a complex collection algorithm should be understood before the data are analyzed. Furthermore, the difficulty of the analysis reside in deciding the proper unit of analysis and predicting the analysis unit ahead of the data collection process. The purpose of this paper is to investigate the potential values of the Internet use data and examine the validity of the Internet use data by comparing server-side data set and client-side data recording from identical Internet users. As a result of the triangulation of two datasets, we could recommend research designs and propose analysis schemes.

Publication Type: Conference Paper/Unpublished Manuscript
Abstract: In this paper, we examine the effects of source bias on statistical inferences drawn from event data analyses. Most event data projects code actors, targets, events, and dates from the same news report. For example most of the early Kansas Event Data System (KEDS) datasets code only Reuters news reports, while recent KEDS datasets code Agence France Presse (AFP) reports. One of the goals of Project Civil Strife (PCS) ?a new event data project? is to code event data from several news sources and examine the effects that source bias has on the inferences we draw from statistical time-series models. In this study, we concentrate on Cambodia from 1980-2005 using automated content analyzed datasets collected from multiple sources (i.e. BBC, Xinhua, Jiji Press Ticker, Associated Press, and AFP). Provided we draw different inferences from different sources, we combine the different datasets into a single comprehensive dataset and eliminate duplicate events to generate a less biased dataset.