Jack Wilkinson
@jdwilko.bsky.social
Centre for Biostatistics, Uni of Manchester. INSPECT-SR, a tool to identify problematic clinical trials. Research misconduct, fertility research. Own opinions.
Trying to reproduce results from old papers can take you to interesting places
September 18, 2025 at 1:36 PM
Trying to reproduce results from old papers can take you to interesting places
If you'd like a preview of INSPECT-SR, well here it is. But it is clear we need a comprehensive (and evolving) guidance document so that people use it properly. We are drafting v1.0 of that to accompany publication. Fingers crossed, we can share the preprint next month! tinyurl.com/cv9muuc4
June 26, 2025 at 12:07 PM
If you'd like a preview of INSPECT-SR, well here it is. But it is clear we need a comprehensive (and evolving) guidance document so that people use it properly. We are drafting v1.0 of that to accompany publication. Fingers crossed, we can share the preprint next month! tinyurl.com/cv9muuc4
Our Delphi will put these questions to experts - ppl who have experience applying for funding for RCTs, pilot, or feasibility studies, and ppl who review these funding bids for funders (as member of a grant panel, or as a reviewer). If that sounds like you - please register to participate (QR code)
March 5, 2025 at 3:48 PM
Our Delphi will put these questions to experts - ppl who have experience applying for funding for RCTs, pilot, or feasibility studies, and ppl who review these funding bids for funders (as member of a grant panel, or as a reviewer). If that sounds like you - please register to participate (QR code)
We have classified markers of promise into the following domains (a work in progress). Questions to answer: which domains are important to demonstrating promise? Are we missing any? What constitutes good evidence for each? And how does this vary at different stages of the research pathway?
March 5, 2025 at 3:48 PM
We have classified markers of promise into the following domains (a work in progress). Questions to answer: which domains are important to demonstrating promise? Are we missing any? What constitutes good evidence for each? And how does this vary at different stages of the research pathway?
We are undertaking a consensus process. We have reviewed funder guidance and 100 protocols to see what is being asked for and what people are supplying as evidence of promise. We will enter these results into a Delphi process to see what the community thinks.
March 5, 2025 at 3:48 PM
We are undertaking a consensus process. We have reviewed funder guidance and 100 protocols to see what is being asked for and what people are supplying as evidence of promise. We will enter these results into a Delphi process to see what the community thinks.
Funders expect evidence of 'promise' (the intervention is likely to work) when considering whether to fund an RCT, pilot, or feasibility study. But there is no shared understanding of how we should do this. We have been funded by @nihr.bsky.social to develop guidance to help.
March 5, 2025 at 3:48 PM
Funders expect evidence of 'promise' (the intervention is likely to work) when considering whether to fund an RCT, pilot, or feasibility study. But there is no shared understanding of how we should do this. We have been funded by @nihr.bsky.social to develop guidance to help.
I’ll bring my stapler
February 13, 2025 at 12:20 PM
I’ll bring my stapler
I was asked for examples. Here are a few pulled from recent cases. It was suggested that there was an error in p-values when recalculated. But in fact these are all consistent (hopefully I have not made a typo). Need to calculate the range of p-values that would be compatible, given rounding.
February 8, 2025 at 2:08 PM
I was asked for examples. Here are a few pulled from recent cases. It was suggested that there was an error in p-values when recalculated. But in fact these are all consistent (hopefully I have not made a typo). Need to calculate the range of p-values that would be compatible, given rounding.
Here are a few examples pulled from recent cases Daniel. For these, it was suggested that there was an inconsistency - but these are all consistent (hopefully I have not made an error in typing these quickly).
February 8, 2025 at 2:05 PM
Here are a few examples pulled from recent cases Daniel. For these, it was suggested that there was an inconsistency - but these are all consistent (hopefully I have not made an error in typing these quickly).
@flavioazevedo.bsky.social there appears to be an account imitating you - not sure if you are aware. I reported, but still there…(need to expand image to see)
January 31, 2025 at 2:41 PM
@flavioazevedo.bsky.social there appears to be an account imitating you - not sure if you are aware. I reported, but still there…(need to expand image to see)
Who doesn’t love a bit of mandatory training
January 30, 2025 at 11:33 AM
Who doesn’t love a bit of mandatory training
Watching Stefan Stender explain just how little effort is required to perform two-sample Mendelian Randomisation. Run this code on publically-available data, and you get results, sensitivity analyses, plots... no thought required.
December 10, 2024 at 7:38 PM
Watching Stefan Stender explain just how little effort is required to perform two-sample Mendelian Randomisation. Run this code on publically-available data, and you get results, sensitivity analyses, plots... no thought required.
Take a look at the summary baseline data from an RCT. This is a baseline variable. We have means and SDs per group, and a p-value of 0.38 on the right.
December 10, 2024 at 4:40 PM
Take a look at the summary baseline data from an RCT. This is a baseline variable. We have means and SDs per group, and a p-value of 0.38 on the right.
October 18, 2024 at 11:30 AM
I’m out of here. Nice time at #ictmc2024 . If you’re interested in viewing the draft INSPECT-SR tool, or signing up to test it, here are the details.
October 3, 2024 at 9:54 AM
I’m out of here. Nice time at #ictmc2024 . If you’re interested in viewing the draft INSPECT-SR tool, or signing up to test it, here are the details.
Or come and find myself or Sarah to get a card.
October 2, 2024 at 9:30 AM
Or come and find myself or Sarah to get a card.
We hope to draw upon a broad range of experience in our Delphi, so please consider registering if you feel you could contribute.
October 2, 2024 at 9:24 AM
We hope to draw upon a broad range of experience in our Delphi, so please consider registering if you feel you could contribute.
We’ve been funded to develop guidance on this - this is how we’re doing it.
October 2, 2024 at 9:22 AM
We’ve been funded to develop guidance on this - this is how we’re doing it.
Funders often require evidence of ‘promise’ (that an intervention is likely to work) before funding a trial or pilot RCT. But there is no agreement on how to do this.
October 2, 2024 at 9:21 AM
Funders often require evidence of ‘promise’ (that an intervention is likely to work) before funding a trial or pilot RCT. But there is no agreement on how to do this.
Great talk from my colleague Sarah Cotterill introducing our Promise study.
October 2, 2024 at 9:20 AM
Great talk from my colleague Sarah Cotterill introducing our Promise study.
First Altman sighting, lfg #ictmc2024
October 1, 2024 at 8:06 AM
First Altman sighting, lfg #ictmc2024
Last few days to sign up for the INSPECT-SR workshop at #ICTMC2024. Come and learn how to use INSPECT-SR to assess trustworthiness of RCTs (hands on), plus learn about some other tools and frameworks (REAPPRAISED, IPD Integrity Tool). ictmc.org/workshop-pro...
September 24, 2024 at 9:27 AM
Last few days to sign up for the INSPECT-SR workshop at #ICTMC2024. Come and learn how to use INSPECT-SR to assess trustworthiness of RCTs (hands on), plus learn about some other tools and frameworks (REAPPRAISED, IPD Integrity Tool). ictmc.org/workshop-pro...
We have arrived in Prague for #GES2024 , where we will talk about problematic clinical trials. We spent the first day committing data duplication (left) and falsification (right).
September 9, 2024 at 4:10 PM
We have arrived in Prague for #GES2024 , where we will talk about problematic clinical trials. We spent the first day committing data duplication (left) and falsification (right).
My colleague Calvin Heal is currently exploring Chernoff faces as a tool for forensic data analysis. Each face represents a multivariate observation in a dataset.
August 31, 2024 at 7:18 AM
My colleague Calvin Heal is currently exploring Chernoff faces as a tool for forensic data analysis. Each face represents a multivariate observation in a dataset.