مقاله انگلیسی رایگان در مورد در مورد اهمیت نتایج آماری در آزمایشات رفتار مصرف کننده – اسپرینگر ۲۰۱۸

Experimentation is the sine qua non of consumer behavior research. To illustrate, in two recent volumes of the Journal of Consumer Research (volumes 40 and 41), of the 179 research articles, 151 (or 84%) reported the results of experiments. Moreover, a majority of the articles reporting experimental research results contain multiple experiments. The prototypical experiment in consumer behavior research consists of (1) specifying (usually null) hypotheses based on some theory, (2) creating an experimental design, (3) implementing the experimental design (including empirical data collection), (4) assessing the hypotheses by means of statistical analysis, and (5) drawing inferences. Statistical analysis commonly involves an analysis of variance (ANOVA) in which a treatment effect—an experimental manipulation—is compared to an estimate of experimental Berror^ by means of an F-statistic. The significance, or lack of significance, of a calculated Fstatistic at some value of p is the fundamental basis used to test the success or failure of an experimental treatment or manipulation. If the F-statistic is significant, the null hypothesis is deemed to be rejected, and the treatment or manipulation is considered to be successful. If the F-statistic is not significant, typically a researcher mentions in passing that it was not significant, moves on to F-tests of other treatments or manipulations in the experiment, and rarely, if ever, discusses the implications of a non-significant treatment or manipulation. An exception to this common treatment was reported by Duclos et al. (2013, p. 130). Although they acknowledged that their two-way ANOVAs did not reveal any significant main or interactive effects (F-statistics were respectively .64, .05, .08, .14, 2.35, and .00 [some of which were significantly insignificant]), they noted that BWhile we could have stopped here, we nonetheless proceeded with a follow-up ANCOVA ….^ Stated somewhat differently, testing a null hypothesis by means of analysis of variance can result in one of the three outcomes. One outcome is that the F-statistic is Bstatistically significant^ in that it is greater than some theoretical value.