Rebecca Sutton Koeser: What do you do when your data has gaps & unknowns? In “Missing Data, Speculative Reading” @zoeleblanc.bsky.social and I delve into Shakespeare & Company datasets to estimate & speculatively fill in gaps. modernismmodernity.org/forums/posts... culturalanalytics.org/article/1169...

What do you do when your data has gaps & unknowns? In “Missing Data, Speculative Reading” @zoeleblanc.bsky.social and I delve into Shakespeare & Company datasets to estimate & speculatively fill in gaps. modernismmodernity.org/forums/posts... culturalanalytics.org/article/1169...

Missing Data, Speculative Reading | Modernism / Modernity Print+modernismmodernity.org Sylvia Beach’s Shakespeare and Company, once a hub for the Lost Generation, has become iconic in popular culture.[1] The continued cultural prominence of the bookshop and lending library is no accident. Beach knew that publishing James Joyce’s Ulysses (1922) would make Shakespeare and Company famous. Her memoir, published in 1959, is an exercise in name-dropping. She

We know from prior work that there are gaps in the archival record (as with most archives!). Missing logbooks in the 1930s, incomplete borrowing records. shakespeareandco.princeton.edu/analysis/202...

The Shakespeare and Company Lending Library Cards in Contextshakespeareandco.princeton.edu What is the relationship between the lending library cards and the logbooks?

We use time series forecasting to estimate missing membership activity (subscriptions, renewals, reimbursements) and members and conclude that Shakespeare and Company Project datasets include 93% of membership activity and 96% of members.

We calculated a “borrowing capacity” based on known subscription limits & discovered members went over subscription limits and Sylvia Beach got more lenient in later years. We compare to a handwritten tally from 1931: 75 books out is 2x our estimate!

We used Copia by Mike Kestemont, @folgertk.bsky.social et al. on Shakespeare and Company lending library books, treating each borrow as a “sighting” of a title. We estimate the data includes 64-76% of books despite only ~26% of borrowing activity.

Speculation! We implemented recommendation systems using known Shakespeare and Company borrowing to predict missing borrowing activity. Hemingway is our test case: his early membership borrowing records don’t survive, lots of documentation to check our results (letters, book catalogs).

@zoeleblanc.bsky.social and I wrote SO much code for this article! 😅 (Not all of it used in this paper.) Replication code/data on @CulturalAnalyt dataverse doi.org/10.7910/DVN/... ; even more code on GitHub github.com/rlskoeser/sh...

Replication Code and Data for: Missing Data, Speculative Readingdoi.org Replication code and data for Missing Data, Speculative Reading Python code, Jupyter notebooks and related data. Applies a combination of time seri...

Post