Week 7: A Look at Power Analysis & Designing a Study

G*Power is a great tool to conduct an a priori analysis to identify the number of participants you might need given values for significance-p, effect size, and power. Download at: https://stats.idre.ucla.edu/other/gpower/

Also, try out this Balanced Latin Squares Calculator

Examples of G*Power:

Step through at least 3 from the following list- I have starred (*) the recommended 3. If you have more time, it may be helpful to you to go through all of them.

Activity- Design an experiment to test:

Choose ONE of the following scenarios and design an experiment. (see detailed descriptions of scenarios below)

Scenario One: A Haptic Device
Scenario Two: A New Authoring Tool, DESIGNX
Scenario Three: Comparing Two Embodiment Conditions
Scenario Four: Investigating Auditory Feedback on a Selection Task
Scenario Five: A Novel Training Platform

In your research notebook, reflect on the following:

Describe your experimental design. Reflect on the research questions and describe a hypothesis. What are the independent variables, dependent variables, tasks, etc?
Will you use a between subjects design, a within subjects design, or a mixed design?
Will you study all independent variables in one experiment or break them up into multiple experiments?
Given your experimental design, what analysis might you choose for each of the dependent variables?
What led you to your choices?
Use G*Power and estimate your tentative participant sample size given various effect sizes: low, medium, high.
What did you learn?
What revisions (if any) would you make based on this power analysis?

Discussion Prompt: Beyond studying new methods and devices-- Can VR be used as a tool to conduct your study?

Read the following paper to facilitate discussion: Designing Immersive Virtual Environments for Human Behavior Research, then add your thoughts to your research notebook.

**Optional** Extra Tidbit: Studying human behavior with virtual reality: The Unity Experiment Framework

------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Details of the Scenarios: (for "Choose ONE of the following scenarios and design an experiment")

Scenario One: A Haptic Device

The purpose of this experiment is to determine whether our novel haptic device HAPTICOGREAT can provide an impact sensation according to our model. We want to measure the user’s subjective responses on impact sensation - (ie whether it feels like hitting a ball). We can also record the pressure applied by our device at the moment the user swings the device to hit an object. We want to understand how well our haptic device provides these sensations.

Scenario Two: A New Authoring Tool, DESIGNX

We designed a brand new authoring tool called DESIGNX. The purpose of our experiment is to understand how well users can complete tasks using DESIGNX, and we want to compare two methods of interaction- DIRECT interaction and INDIRECT. For Direct, users will use their virtual hands to complete tasks and for INDIRECT an alternative method will be provided for them. These are two design considerations for how users can use DESIGNX capabilities. We want to investigate for example how fast or how accurate to complete a task, how fast to navigate, how fast to select, which method is preferred, etc.

Scenario Three: Comparing Two Embodiment Conditions

The purpose of this study is to explore the influence of embodiment on body weight perception of a virtual human while keeping the degree of its personalization constant.We want to compare body weight estimations comparing our new embodiment illusion condition EMBODY to a no embodiment illusion condition NOBODY. In the EMBODY condition, a mirror is provided for the user to see themselves. In the NOBODY condition, the virtual mirror is replaced by a door frame leading to another room where the user can see another virtual human perform the same movements. We will record the actual BMI of each participant and the estimated BMI of the virtual human where the estimate is completed by the participants for EMBODY / NOBODY. We want to know how the use of the EMBODY condition compares when participants estimate the BMI.

Scenario Four: Investigating Auditory Feedback on a Selection Task

The purpose of this study is to investigate how different sounds as error feedback during a TARGET SELECTION task (move one object from A to B) affect user performance and with different task completion strategies. We have two different auditory feedback types. We also can compare using two different types of common selection techniques - RAY_CAST (using a ray to select) and VIRTUAL_HAND (using a finger point or hand to select). We can ask participants to select a presented target as fast and as accurate as possible. We want to know how the auditory feedback types play a role in the user’s performance on those tasks.

Scenario Five: A Novel Training Platform

The purpose of this study is to investigate whether participants’ motor-task performance in VR improves after a two-weeks training period in our simulator SIMX, and whether this improvement is influenced by having their hands represented in the VE during training. SIMX will provide a simulation of a bi-manual tool-based pick and place task. We want to compare our SIMX training, and ask would training groups significantly improve their performance in the VR simulators as compared to the control group? In our SIMX training we can provide the visualization of the two hands or just the tools, and wonder if that makes a difference.