Research Proposal, Design, and G*Power Principles

Research Proposal Development

Next Wednesday's class will be dedicated to selecting a research topic for the proposal. We will brainstorm approximately $12$ to $14$ different topics, which will then be narrowed down to the "least worst option." This is because finding a research question that universally excites and appeals to everyone is unlikely. The chosen topic must be amenable to a single-factor experimental design.

Today, we aim to cover power and effect size. Anything not finished today will be completed in tomorrow's lab session, which is expected to be short (around $45$ minutes at most for G*Power demonstration). It is crucial to start thinking about solid research questions now, rather than waiting until the last minute before class on Wednesday. Consider questions you genuinely want to explore and potentially build your research proposal around.

Generating Research Questions

Several strategies can help in formulating strong research questions:

1. Build from Existing Research: Look at the "Future Research" sections in scholarly articles or within chapters that discuss related studies. If a particular topic interests you, consider how you can expand upon or build off those existing studies. For instance, in a psychology class, delving into a chapter on object and face recognition helped brainstorm ideas.

2. Everyday Experiences and Genuine Interests: As psychology students, you are expected to be observant of human behavior. Your own daily experiences and genuine curiosities can spark compelling research questions.

   • Example: Cues of Attentiveness: A personal experience of feeling ignored led to the question: "What type of cues do I rely on to infer whether someone is paying attention to me or not? Are there any cues of attentiveness during a conversation?" This evolved into a graduate student project investigating pupil size as a cue for judging attentiveness in others. The rationale was that people spend considerable time staring at each other's eyes during conversations (often unconsciously), and changes in pupil diameter are detectable at an arm's length (a typical conversation distance). Pupil size is known to correlate with working memory and cognitive load, suggesting that if someone is focused in a conversation, their pupil size likely changes.

Examples of Potential Research Topics

Aside from attentiveness, other areas of interest mentioned include:

• Positive Emotion: Investigating why certain activities lead to positive emotion and identifying the factors that influence this phenomenon.
• Mind-Wandering: Exploring its causes, influencing factors, predispositions in certain individuals, typical content, and duration.
• Artfulness and Sense of Thought: (Although vague, it exemplifies exploring an everyday experience that one wishes to understand better).
• Functions of Emotions: While negative emotions (e.g., fear for danger, disgust for preventing ingestion, sadness for seeking support) have clear evolutionary functions, the functions of positive emotions are often less clear, presenting an interesting area for research.
• Preference for Curved Architecture: Observing that people tend to prefer curved architecture, find it more aesthetically pleasing, and are willing to pay more for it compared to angular designs.
  • Explanation 1: A natural, ingrained preference or disposition for curved shapes.
  • Explanation 2: Angular shapes might be subconsciously associated with danger (e.g., sharp teeth, thorns, sharp rocks). This association triggers a "population of effect," leading to a decreased preference for angular forms.

Understanding Experimental Design Principles

• Within-Subjects Design: This design re-exposes the same individual to different conditions. Its primary advantage is eliminating alternative explanations for observed differences that might arise from subject variability. For example:
  • Differences in personality can't be an alternative explanation because the same person's personality doesn't drastically change between testing sessions.
  • Differences in intelligence are also ruled out, as intelligence doesn't fluctuate significantly in this manner.
    By controlling for individual differences, a within-subjects design increases the likelihood of detecting an effect. If a study manipulates more than one factor (e.g., through an interaction effect, as discussed in factorial designs), there are more potential reasons (manipulation, interaction) for an observed effect.
• Tight Experimentation: This involves meticulously controlling experimental conditions. For example, ensuring that everyone in a study group is exposed to the same drug, tested at the same time of day, and given identical instructions. The goal is to minimize confounding variables, ensuring that any differences in outcomes are solely attributable to the independent variable being manipulated (e.g., the drug).

The Importance of G*Power and Effect Size

G*Power is a critical tool for experimental design, particularly in considering factors that influence the outcome of null hypothesis testing. The alpha level ($\alpha$), or Type I error rate, plays a significant role:

• Increasing $\alpha$ (e.g., from $0.01$ to $0.05$) increases the probability of falsely rejecting the null hypothesis (i.e., finding an effect when there isn't one). At $\alpha = 0.05$, there's a $5\%$ chance that your data could show a significant difference even if it comes from a population with no real effect. This means $5$ out of $100$ times, you might observe a large difference between samples that actually originate from the same population, potentially wasting time and resources.
• If a study has a low probability of finding a genuine effect, it should be redesigned to increase its chances of success before execution.
• G*Power helps researchers account for multiple factors (including alpha, effect size, sample size, and power) to design studies that are likely to detect meaningful effects.

Statistical vs. Practical Significance

It is crucial to differentiate between statistical significance and practical significance.

• Sample Size Influence: The outcome of null hypothesis testing is always influenced by sample size. By collecting enough data, one can almost always find statistically significant differences between groups, even if those differences are trivial in the real world.
• Trivial Differences Example: If you tested $10,000$ males and $10,000$ females, you might statistically demonstrate a sex difference in intelligence. However, this difference could be practically negligible, perhaps as small as $0.0001$ of an IQ unit. While statistically significant due to the large sample size, it holds no practical importance.
• This highlights why tools like G*Power are essential: they help researchers design studies that not only seek statistical significance but also aim to uncover effects that are meaningful and substantial in practical terms.