Measurement Bias in Multilevel Data
In: Structural equation modeling: a multidisciplinary journal, Band 21, Heft 1, S. 31-39
ISSN: 1532-8007
1521 Ergebnisse
Sortierung:
In: Structural equation modeling: a multidisciplinary journal, Band 21, Heft 1, S. 31-39
ISSN: 1532-8007
In: Technical reports 64
In: Structural equation modeling: a multidisciplinary journal, Band 19, Heft 4, S. 561-579
ISSN: 1532-8007
In: The American review of public administration: ARPA, Band 45, Heft 5, S. 542-564
ISSN: 1552-3357
The study of managerial networking has been growing in the field of public administration; a field that analyzes how managers in open system organizations interact with different external actors and organizations. Coincident with this interest in managerial networking is the use of self-reported survey data to measure managerial behavior in building and maintaining networks. One predominant approach is to generate factor indices of networking activity from ordinal scales. However, when public managers answer survey questions with ordinal scales to describe their networking activities, the answers may be subject to various response biases. Consequently, the use of factor indices may lead to biased measurements that misrepresent managerial networking. As an alternative, we build on studies that apply the item response theory (IRT) as a measurement strategy and propose a Bayesian alternative. To tap managers' latent effort put in networking activity, the Bayesian Generalized Partial Credit Model allows us to select a one-dimensional networking scale from multiple ordinal survey items. Using 12 such items in a mail survey of nearly 1,000 American hospital managers, we demonstrate the advantage of using the Bayesian IRT model over factor-analytic models in a substantive test of how managerial networking affects organizational performance.
SSRN
In: American review of public administration: ARPA, Band 45, Heft 5, S. 542
ISSN: 0275-0740
In: Journal of aging studies, Band 7, Heft 4, S. 453-464
ISSN: 1879-193X
In: Working paper series 47
SSRN
Working paper
In: AWWA water science, Band 1, Heft 2
ISSN: 2577-8161
Two chlorinated cyanurates, commonly referred to as dichlor (anhydrous sodium dichloroisocyanurate or sodium dichloroisocyanurate dihydrate) and trichlor (trichloroisocyanuric acid), may be approved for use in U.S. drinking water systems as chlorine sources. One complication with dichlor or trichlor's application in drinking water is that the actual free chlorine concentration in these systems cannot be quantified accurately by currently approved methods. Based on known water chemistry, two hypothesized advantages of dichlor or trichlor use are potential increased residual chlorine stability and decreased regulated disinfectant byproduct (DBP) formation. To inform these practical considerations, the current research investigated measurement bias in N,N‐diethyl‐p‐phenylenediamine (colorimetric and portable parallel analyzer), indophenol, amperometric titration, and amperometric electrode free chlorine methods. In addition, hold studies using a surface water and dosed with either free chlorine only, dichlor, or trichlor provided the first side‐by‐side comparisons of disinfectant residual stability and regulated DBP formation.
In: Journal of international economics, Band 87, Heft 1, S. 105-111
ISSN: 0022-1996
In: Survey methods: insights from the field
ISSN: 2296-4754
In 2011, a large-scale mixed-mode experiment was linked to the Crime Victimisation Survey (CVS). This experiment consisted of a randomized allocation of sample persons to the four contemporary survey modes Web, mail, telephone and face-to-face, and a follow-up using only interviewer modes. The aim of the experiment was to disentangle mode-specific selection- and measurement bias. In a previous paper (Schouten et al 2013), mode-specific selection and measurement biases were reported for a large number of key variables from the CVS and the Labour Force Survey. This paper is a follow-up to that study and investigates the size of the selection and measurement biases as a function of contact effort, where contact effort refers to the number of telephone calls, the number of face-to-face visits and the number of reminders in Web and mail. In the analyses, face-to-face response based on a maximum of six visits, the default face-to-face strategy at Statistics Netherlands, is used as the benchmark.
The analyses show that contact effort has little impact on the size of measurement bias and a modest impact on the size of selection bias. From the results, it is therefore concluded that contact effort is not a strong common cause of nonresponse and measurement error.
In: Research policy: policy, management and economic studies of science, technology and innovation, Band 53, Heft 5, S. 104992
ISSN: 1873-7625
In: Journal of survey statistics and methodology: JSSAM, Band 5, Heft 4, S. 409-432
ISSN: 2325-0992
In: Chinese political science review, Band 1, Heft 2, S. 268-302
ISSN: 2365-4252