Goldilocks and the (approximately) Just Right Significance Level by Sarah Hartman ’25

Goldilocks and the (approximately) Just Right Significance Level: Estimating False Discovery Rates for Genome-wide Studies by Sarah Hartman ’25, Wednesday February 26, 1:00 – 1:50pm, North Science Building 015, Wachenheim, Statistics Colloquium

Goldilocks (the scientist!) seeks to discover genes that are expressed differently between two different types of cancer cells. With advanced technology, such as microarrays and RNA-sequencing, she can conduct hypothesis tests for THOUSANDS of different genes at once. She wants to identify which genes are “statistically significant”, however the traditional scientifically accepted p-value cutoff of .05 is TOO BIG, as it can unfortunately result in hundreds of false discoveries. However, if she chooses a p-value cutoff that is TOO SMALL, she may leave many genes undiscovered. In this case, a useful measure of significance that balances limiting error with preserving statistical power is the False Discovery Rate, the expected proportion of false positives among the genes that are discovered. We will use the exciting properties of p-values to estimate the False Discovery Rate, which can be used in a wide variety of statistical and scientific applications!