Episode 11 - Replication crisis and the importance of open science practices

Tonight (episode 11), before diving into statistical analysis with R in the coming week, I explained the replication crisis and the importance of open science practices like pre-registration. In the first part, I discussed why so many research findings fail to replicate and explored the underlying causes. In the second part, I focused on what we can do to improve reproducibility and how to implement these solutions. I concluded the lesson by demonstrating how to adopt open science practices using platforms like OSF.

Lab 9 and 10 – Power analysis with R and the importance of novelty in research
Deadline: Tuesday, 31st December 2024 (5pm Nigeria time)

1. (a) What does power represent in power analysis? (b) what are the consequences of conducting an underpowered study? (c) What are the components of power analysis?

2. Use the pwr package in R or G*Power software to determine the sample size required for detecting a medium effect size (Cohen’s d = 0.5) with a significance level of 0.05 and a power of 0.8. Submit a screenshot of your code or software output along with an explanation of the parameters and results. Finally, compare your results from R with those obtained using G*Power. What similarities or differences do you notice?

3. Ensure that your research question is novel or includes a novel element. Start by using academic search engines such as Google Scholar, PubMed, or Scopus to verify whether your research question or topic has been addressed before. Can you identify gaps or areas where novelty can be added, such as applying the concept in a new context or population or using a different methodology? Document your findings by noting whether your research question already exists and, if it does, what novel element you will introduce. Based on these findings, refine your research question to include the identified novel element. Clearly state how your approach contributes something new or unique to existing knowledge. Document a short report that includes your original research question, the results of your novelty check (including similar studies if found), and your revised research question with the added novelty.

In the last episode (Episode 10), I covered the topic of power analysis and demonstrated how to carry it out using the pwr package in R. Power analysis is crucial for determining the required sample size to detect an effect if it exists, ensuring that a study is neither overpowered (wasting resources) nor underpowered (risking Type II error).

I explained key concepts and terminologies, such as power, effect, effect size, sample size, and level of significance. In addition to using R, I also demonstrated how to perform power analysis using the G*Power software.

Today (episode 9), as we move closer to the stage of getting data that we are going to ultimately analyze with R, I continued my explanation following our previous discussion on using Randomized Controlled Trials (RCTs) in the last session. I emphasized the importance of novelty in research, explained why it matters, and outlined steps to check for novelty effectively before carrying out the research work.

I concluded the lesson by discussing the importance of pilot studies in research and the potential consequences of skipping this critical step. I illustrated the significance of piloting interventions with three examples and Nigeria as a case study: one related to the 2023 election and two recent policies introduced by the current government.

Episode 5

I introduced packages. So far, we have used functions available in the base distribution of R. Since R is open source, we may want to use functions written by others for our analyses. A common and useful package for data manipulation is tidyverse, which includes tools like dplyr (for data manipulation) and ggplot2 (for data visualization). Tonight, I focused on dplyr and explained some of its useful functions for data manipulation, such as filter(), select(), group_by(), summarise(), and arrange().