http://arxiv.org/pdf/hep-th/9610145.pdf
[PDF]Decision making with incomplete information - LUISS.it
static.luiss.it/hey/ambiguity/papers/Weber_1987.pdf
by M WEBER - 1987 - Cited by 357 - Related articles
Abstract: Decision situations with incomplete information are characterized by a decision maker without a ... consumer preferences such as conjoint analysis.An Overview and Comparison of Design Strategies for Choice-Based Conjoint Analysis
Keith Chrzan, Maritz Marketing Research Bryan Orme, Sawtooth Software
There are several different approaches to designing choice-based conjoint experiments and several kinds of effects one might want to model and quantify in such experiments. The approaches differ in terms of which effects they can capture and in how efficiently they do so. No single design approach is clearly superior in all circumstances.
This paper describes different kinds of design formats (full profile, partial profile), and different methods for making designs (manual, computer optimization, computer randomization) for choice-based conjoint designs. Over and above the plain vanilla generic main effects most commonly modeled in conjoint analysis, there are several types of “special effects” that can be included in choice-based models. The various ways of constructing choice-based designs are compared in terms of their ability to capture these effects. Using simulations and artificial data sets we also assess the statistical efficiency of the various design methods.
An Overview and Comparison of Design Strategies for Choice-Based Conjoint Analysis
Keith Chrzan, Maritz Marketing Research Bryan Orme, Sawtooth Software
There are several different approaches to designing choice-based conjoint experiments and several kinds of effects one might want to model and quantify in such experiments. The approaches differ in terms of which effects they can capture and in how efficiently they do so. No single design approach is clearly superior in all circumstances.
This paper describes different kinds of design formats (full profile, partial profile), and different methods for making designs (manual, computer optimization, computer randomization) for choice-based conjoint designs. Over and above the plain vanilla generic main effects most commonly modeled in conjoint analysis, there are several types of “special effects” that can be included in choice-based models. The various ways of constructing choice-based designs are compared in terms of their ability to capture these effects. Using simulations and artificial data sets we also assess the statistical efficiency of the various design methods.
Background
In traditional conjoint analysis (see Figure 1), experimentally controlled combinations of attribute levels called profiles are presented to respondents for evaluation (ratings or rankings). In a multiple regression analysis these evaluations then become the dependent variables predicted as a function of the experimental design variables manifested in the profiles.
2
In 1983, however, Louviere and Woodworth extended conjoint analysis thinking to choice evaluations and multinomial logit analysis. In the choice-based world, respondents choose among sets of experimentally controlled sets of profiles and these choices are modeled via multinomial logit as a function of the experimental design variables.
As you might guess, the greater complexity of the experiment allows the researcher to think about designing and estimating many more interesting effects than the simple main
Panasonic Micron filtration 12 amps 3 year warranty $149
Rate: 1 2 3 4 5 6 7
Attribute levels
A profile
Experimental design
Figure 1 - Traditional Ratings-Based Conjoint
Panasonic Micron filtration 12 amps 3 year warranty 12 ft cord $149
Eureka Micron filtration 8 amps 4 year warranty 16 ft cord $129
Hoover Wind Tunnel 10 amps 2 year warranty 18 ft cord $179
A choice
Experimental design
Pick one
Figure 2 - Choice-Based Conjoint Experiment
3
effects and occasional interaction effects of traditional conjoint analysis (Louviere 1988, Anderson and Wiley 1992, Lazari and Anderson 1994).
In addition to focusing on the novel effects choice-based analysis allowed, other topics became important for choice-based analysis. Design efficiency became a topic of research because the efficiency of experimental designs for multinomial logit was not as straightforward as that for traditional linear models and their designs (Kuhfeld et al. 1994, Bunch et al. 1994, Huber and Zwerina 1995). Finally, still other researchers sought ways to make choice-based experiments easier for researchers to design (Sawtooth Software 1999) or for respondents to complete (Chrzan and Elrod 1995).
----------I. INTRODUCTION The concept of interacting field introduced by Faraday-Maxwell is on the basis of Field Theory and of Quantum Mechanics. It is necessary for a relativistic description of the wave-particle duality of the quantum theories. The ide a of a classical interaction as a wave, continuous and distributed over the whole space, is put against the modern idea of a quantum interaction, discrete and localized in “corpuscles” or interaction quanta. The passage from the first to the second idea requires a so called quantization process, and this passage, it is well known, in the best of the cases (QED) has deep problems, and in the worst case (QG) it has proved to be not viable. In the search of solutions to this problem of quantizing the gravitational field the observed tendency is the one of replacing complex models and formalisms for others of increasing complexity. On the other hand, the Classical Electrodynamics, the best known and “well succeeded” paradigm of all subsequent field theory, has old and well known problems of inconsistency with the description of fields in the neighborhood of their sources. In a recent work (hep-th/9610028) we consider these problems and we show that taking the correct zero distance limit to the charge reveals unequivocal clues of quantum features in Classical Electrodynamics: the flux of field from a charge is discontinuous in time. Out of this limit this discontinuity is masqueraded by the field spacetime-average character. The problems of CED in its zero distance limit, which are erroneously attributed to the working hypothesis of a pointlike electron, are rather unequivocal sign s that our concepts of fields are unappropriate for describing interactions. The idea of a continuous classical field is misleading; it is valid only for large distances and large number of photons. We want to make here a critical review of the Faraday-Maxwell concept of classical field under the perspective of modern physics that understands it as being of a fundamentally quantum nature. It is well known that the pioneers of classical field theory worked with a model for the electromagnetic phenomena based on an analogy with the fluid mechanics. The electromagnetic effects would be propagated through an all-pervading fluid, the ether. This, at the time, new vision of field-mediated interactions between distant charges was an advance with respect to the Newtonian concept of action at a distance. The fluid analogy implies on a field distributed all over the space around its source, like it happens with the sound waves, for example. This image was certainly reinforced by what at that time seemed to be an apparently definitive victory of the concept of light as a wave phenomena against the Newtonian model of light as a stream of corpuscles. The phenomena of interference, diffraction and polarization of light had been decisive for this conviction. Only much later the first clues of a discreteness, like the photoelectric and the Compton effects, would be discovered. But even Quantum Mechanics that was created from the necessity of explaining this new kind of effects received also the influence of this fluid-mechanics concept of field: the wave function is a space-distributed field representing a fluid of amplitude of probabilities. In this qualitative analysis we want to oppose this historical vision of classical fields as some waves, continuous and distributed all over the space, against the vision of interactions mediated by localized point-like objects, their quanta, discretely emitted, propagated and absorbed. In the modern perspective all the four fundamental interactions of nature are mediated by their respective quanta. We will try to obtain a qualitative view on how a classical field theory could be formulated if the interaction were seen as mediated by massless point-like objects propagating on straight-line trajectories between their emitters and their absorbers .
No comments:
Post a Comment