China
Africa
Explore chapters and articles related to this topic
Research Methods
Published in Nancy J. Stone, Chaparro Alex, Joseph R. Keebler, Barbara S. Chaparro, Daniel S. McConnell, Introduction to Human Factors, 2017
Nancy J. Stone, Chaparro Alex, Joseph R. Keebler, Barbara S. Chaparro, Daniel S. McConnell
Although random sampling deals with ensuring that the sample is reflective of the population, there is an even more important consideration when assigning participants to various experimental conditions. This is referred to as random assignment. Random assignment occurs when we use a method such as flipping a coin, rolling a die, or using a computer-generated randomized list that determines to which experimental group or condition the participants will be assigned. The rationale is that students have equal probability of being placed in any of the experimental groups. If you are studying driving behavior in which some individuals are driving with a cell phone and others are driving without a cell phone, the sample of students in the “driving with” group that represents the population is expected to be similar to the sample of students in the “driving without” group. When the groups are assumed to be similar at the beginning of the experiment, we can argue that differences in driving performance are due to the experimenter’s manipulation; otherwise, the cause of our results would be unclear.
The Analysis of Variance for Designed Experiments
Published in William M. Mendenhall, Terry L. Sincich, Statistics for Engineering and the Sciences, 2016
William M. Mendenhall, Terry L. Sincich
As we learned in Chapter 13, the most common assignment of treatments to experimental units is called a completely randomized design. To illustrate, suppose we wish to obtain equal amounts of information on the mean assembly times for the three training procedures; i.e., we decide to assign equal numbers of workers to each of the three training programs. Also, suppose we determine the number of workers in each of the three samples to be n1 = n2 = n3 = 10. Then a completely randomized design is one in which the n1 = n2 = n3 = 30 workers are randomly assigned, 10 to each of the three treatments. A random assignment is one in which any one assignment is as probable as any other. This eliminates the possibility of bias that might occur if the workers were assigned in some systematic manner. For example, a systematic assignment might accidentally assign most of the manually dexterous workers to training program A, thus underestimating the true mean assembly time corresponding to A.
Research Methods in Human Factors
Published in Robert W. Proctor, Van Zandt Trisha, Human Factors in Simple and Complex Systems, 2018
Robert W. Proctor, Van Zandt Trisha
Which particular independent and dependent variables we examine will depend on the hypotheses under consideration. With a random assignment, each person has an equal probability of being assigned to any condition. Random assignment ensures that there can be no systematic influence from extraneous factors, such as education or socioeconomic status, on the dependent variables. Consequently, we can attribute differences among treatment conditions solely to the manipulation of the independent variable. As such, we can make a causal statement about the relation between the independent and dependent variables.
Different Principles of Information Design for Three Online Financial Tasks: Localization, Comparison, and Subtraction
Published in International Journal of Human–Computer Interaction, 2023
Héctor R. Ponce, Richard E. Mayer, Sandra F. Torres
Participants were randomly assigned to one of the four conditions and tested individually. Random assignment was accomplished by a computer-based random algorithm. Each participant was asked to sit in the testing room in front of a computer screen. The experimenter explained the aims of the study. Participants were asked to read and sign the consent form. Each participant was asked to practice with the training material for about 5 min before the experiment began. After the training session, the experimenter gave the instructions, which asked the participant to respond to each question as accurately as possible in the minimum possible time. Participants were asked to compute and answer with absolute differences for subtraction questions. Questions regarding the experimental material were not allowed, except for specific difficulties in using the respective interface. Phones, calculators, paper, and pencils were not allowed during the study. We also tracked eye movements using a Tobii eye-tracking system, but these data are not included in this report in the interests of focusing on the key research questions.
Effects of an aerobic fitness test on short- and long-term memory in elementary-aged children
Published in Journal of Sports Sciences, 2020
Jennifer L. Etnier, Paul M. Sprick, Jeffrey D. Labban, Chia-Hao Shih, Stephen M. Glass, Jarod C. Vance
Descriptive data are presented in Table 1. Fitness was operationalized as mile run time for the 4th and 6th graders and as number of laps for the 2nd graders. Performance data from the runs are presented for descriptive purposes, and these data were converted to z-scores relative to each grade level. Several analyses were conducted to confirm that random assignment had resulted in equivalent groups with regards to relevant variables. Chi-square analyses were used to test the distribution of boys and girls across conditions. One-way analyses of variance (ANOVAs) were conducted to test for differences in age, BMI, and fitness z-scores as a function of condition (comparison, treatment). Lastly, a one-way ANOVA was used to confirm that age differed significantly as a function of grade.
The Generativity of Social Media: Opportunities, Challenges, and Guidelines for Conducting Experimental Research
Published in International Journal of Human–Computer Interaction, 2018
In conducting either a framed or free-simulation experiment, we recommend using stratified random sampling (guideline #9). In other words, we suggest stratifying the whole subject pool based on a variable that researchers expect is most likely to create social heterogeneity within the group to sample from, and then randomly selecting subjects from the created strata. One important way that the MyTable users were systematically different from one another was in terms of their access to informational resources generated by their friends, for some were very well-connected while others were not (they had few friends). Thus, network size (i.e., number of online friends on MyTable) was used as the stratification variable. To implement this approach, we divided the whole user base in six strata according to network size, and then drew members from each stratum to assign to the different groups. ANOVA test results indicated that there was no significant difference between groups on demographic attributes and, most importantly, on network size; thus, this technique was deemed effective to assemble homogenous groups. Interestingly, stratification was useful to both Study 1 and Study 2, but for different reasons. In conducting a framed experiment such as Study 1, a researcher wants to ensure that the groups of individuals who are assigned to different treatments only differ by being exposed to different stimuli, and that they are in theory identical in all other respects. Randomization facilitates this objective because it helps to derive clean estimates of causal effects by taking care of the unobserved heterogeneity that tend to impair the sharpness of estimates in observational settings (in which subjects are not randomly assigned to selected conditions) (Falk & Heckman, 2009). But while random assignment reduces the possibility that groups are inequivalent, it does not ensure it; that is, groups may still showcase differences on certain attributes. An effective means to counteract this possibility is to couple randomization with stratification. Different from Study 1, Study 2 benefited from stratification because randomly sampling from the strata in the panel facilitated obtaining a sample in which there was a high degree of variance in a key marker of subjects’ social capital—network size (from no friends to 25 friends).