Sean Mackinnon
@seanpmackinnon.bsky.social
Instructor at Dalhousie University. Personality, statistics, mixed methods
My mistake, thank you. Deleted the post.
October 28, 2025 at 9:51 PM
My mistake, thank you. Deleted the post.
I've found it increasingly hard to care about some of the more common QRPs like p-hacking when generative AI models can like, entirely fabricate a paper and data from scratch at scale.
I guess completely removing the human from all decisions was the natural end state of focusing on QRPs though
I guess completely removing the human from all decisions was the natural end state of focusing on QRPs though
October 27, 2025 at 6:48 PM
I've found it increasingly hard to care about some of the more common QRPs like p-hacking when generative AI models can like, entirely fabricate a paper and data from scratch at scale.
I guess completely removing the human from all decisions was the natural end state of focusing on QRPs though
I guess completely removing the human from all decisions was the natural end state of focusing on QRPs though
I would say that it's more that using AI models replaces what was formerly a google search on SO or other sites formerly.
So, I would say that the coding is of similar quality, but I get more stuff that I never taught with less understanding. But not enough to be a moral panic for grad students.
So, I would say that the coding is of similar quality, but I get more stuff that I never taught with less understanding. But not enough to be a moral panic for grad students.
October 27, 2025 at 3:17 PM
I would say that it's more that using AI models replaces what was formerly a google search on SO or other sites formerly.
So, I would say that the coding is of similar quality, but I get more stuff that I never taught with less understanding. But not enough to be a moral panic for grad students.
So, I would say that the coding is of similar quality, but I get more stuff that I never taught with less understanding. But not enough to be a moral panic for grad students.
As an instructor of grad students (teaching R), they're still attempting stuff on their own but sometimes getting AI models to do things when they can't figure it out. Code mostly works, but basic understanding of what the code does is a bit lower.
The difference is not as dramatic as I thought
The difference is not as dramatic as I thought
October 27, 2025 at 3:13 PM
As an instructor of grad students (teaching R), they're still attempting stuff on their own but sometimes getting AI models to do things when they can't figure it out. Code mostly works, but basic understanding of what the code does is a bit lower.
The difference is not as dramatic as I thought
The difference is not as dramatic as I thought
That's savage lmao
October 25, 2025 at 9:07 PM
That's savage lmao
Reposted by Sean Mackinnon
You want sources? Here you go. Read the filings yourself, you filthy animals:
Notice of 2024 Annual Meeting & Proxy Statement: s204.q4cdn.com/701424631/fi...
Form 10-K SEC Annual Report: s204.q4cdn.com/701424631/fi...
Don't fuck with me. I'll spend another three hours in Excel. I don't care.
Notice of 2024 Annual Meeting & Proxy Statement: s204.q4cdn.com/701424631/fi...
Form 10-K SEC Annual Report: s204.q4cdn.com/701424631/fi...
Don't fuck with me. I'll spend another three hours in Excel. I don't care.
May 23, 2025 at 5:42 PM
You want sources? Here you go. Read the filings yourself, you filthy animals:
Notice of 2024 Annual Meeting & Proxy Statement: s204.q4cdn.com/701424631/fi...
Form 10-K SEC Annual Report: s204.q4cdn.com/701424631/fi...
Don't fuck with me. I'll spend another three hours in Excel. I don't care.
Notice of 2024 Annual Meeting & Proxy Statement: s204.q4cdn.com/701424631/fi...
Form 10-K SEC Annual Report: s204.q4cdn.com/701424631/fi...
Don't fuck with me. I'll spend another three hours in Excel. I don't care.
I'd juat say collect extra data and use it. Having estimates more precise than planned is more a bonus, than a downside.
If I was doing prereg, I would plan for "Minimum of 100, but data collection continues until Feb 22" (or whatever date).
No optional stopping then.
If I was doing prereg, I would plan for "Minimum of 100, but data collection continues until Feb 22" (or whatever date).
No optional stopping then.
October 24, 2025 at 3:52 PM
I'd juat say collect extra data and use it. Having estimates more precise than planned is more a bonus, than a downside.
If I was doing prereg, I would plan for "Minimum of 100, but data collection continues until Feb 22" (or whatever date).
No optional stopping then.
If I was doing prereg, I would plan for "Minimum of 100, but data collection continues until Feb 22" (or whatever date).
No optional stopping then.
This meltdown over the ad is extremely funny to me.
Like Reagan really was pro free-trade. He also sucked super bad, but the issue is that they can't admit that conservatives can like, believe different things and this is really funny as a result.
Like Reagan really was pro free-trade. He also sucked super bad, but the issue is that they can't admit that conservatives can like, believe different things and this is really funny as a result.
October 24, 2025 at 3:00 PM
This meltdown over the ad is extremely funny to me.
Like Reagan really was pro free-trade. He also sucked super bad, but the issue is that they can't admit that conservatives can like, believe different things and this is really funny as a result.
Like Reagan really was pro free-trade. He also sucked super bad, but the issue is that they can't admit that conservatives can like, believe different things and this is really funny as a result.
I guess they can lick boots all the way until the conservatives are elected and completely defund them
October 23, 2025 at 4:16 PM
I guess they can lick boots all the way until the conservatives are elected and completely defund them
It's like the corgi of horses, A++ would look again
October 22, 2025 at 4:25 PM
It's like the corgi of horses, A++ would look again
I'm sorry, this is deeply upsetting. It's like a creepypasta, but for R code
October 20, 2025 at 1:27 PM
I'm sorry, this is deeply upsetting. It's like a creepypasta, but for R code
Whoaaa one guy did all those movie covers and the cover for sabbath bloody sabbath? TIL, and incredible
October 16, 2025 at 9:59 PM
Whoaaa one guy did all those movie covers and the cover for sabbath bloody sabbath? TIL, and incredible
The whole thing is bad clearly, but I'm hung up on why they want a 0 to 42 scale for pain
Like, why?
Like, why?
October 15, 2025 at 8:19 PM
The whole thing is bad clearly, but I'm hung up on why they want a 0 to 42 scale for pain
Like, why?
Like, why?
Ah, I often make my students do it. Did it in this paper:
psycnet.apa.org/record/2023-...
I've never like, put residual plots in a Word document supplement before, but the R code could produce them.
Word counts often mean it ends up on the cutting room floor in papers, even though I do it, tbh.
psycnet.apa.org/record/2023-...
I've never like, put residual plots in a Word document supplement before, but the R code could produce them.
Word counts often mean it ends up on the cutting room floor in papers, even though I do it, tbh.
October 15, 2025 at 1:51 PM
Ah, I often make my students do it. Did it in this paper:
psycnet.apa.org/record/2023-...
I've never like, put residual plots in a Word document supplement before, but the R code could produce them.
Word counts often mean it ends up on the cutting room floor in papers, even though I do it, tbh.
psycnet.apa.org/record/2023-...
I've never like, put residual plots in a Word document supplement before, but the R code could produce them.
Word counts often mean it ends up on the cutting room floor in papers, even though I do it, tbh.
If your TV is very well secured and can't be knocked over, try looking up TV for cats on youtube or wherever. Our cat went wild for it, but we worried she would knock over the TV to she was so into it
October 14, 2025 at 11:26 AM
If your TV is very well secured and can't be knocked over, try looking up TV for cats on youtube or wherever. Our cat went wild for it, but we worried she would knock over the TV to she was so into it
While the old adage holds some truth, gotta be at least some astroturfing going on if the hate mail is internet based, I would imagine (or at least, hope)
October 13, 2025 at 8:58 PM
While the old adage holds some truth, gotta be at least some astroturfing going on if the hate mail is internet based, I would imagine (or at least, hope)
Yeah, I agree with that. I first get into it when teaching multiple regression explaining what all the slopes mean in words, so different context i think
October 13, 2025 at 2:59 PM
Yeah, I agree with that. I first get into it when teaching multiple regression explaining what all the slopes mean in words, so different context i think
When I teach it, I get them to check but that it can't be divorced from the semantic meaning of the two covariates.
In a world where everyone has a clear head in terms of causality and confounding from the study outset, sure it doesn't really matter but students are often forging ahead w/o that imo
In a world where everyone has a clear head in terms of causality and confounding from the study outset, sure it doesn't really matter but students are often forging ahead w/o that imo
October 13, 2025 at 2:33 PM
When I teach it, I get them to check but that it can't be divorced from the semantic meaning of the two covariates.
In a world where everyone has a clear head in terms of causality and confounding from the study outset, sure it doesn't really matter but students are often forging ahead w/o that imo
In a world where everyone has a clear head in terms of causality and confounding from the study outset, sure it doesn't really matter but students are often forging ahead w/o that imo
I think people talk around each other on this, b/c if its
y ~ c1 + c2 + x1
Colinearity among c1 & c2 doesn't really hurt the slope for x, but it does often negatively affect the interpretation of the slopes for c1 & c2.
y ~ c1 + c2 + x1
Colinearity among c1 & c2 doesn't really hurt the slope for x, but it does often negatively affect the interpretation of the slopes for c1 & c2.
October 13, 2025 at 2:30 PM
I think people talk around each other on this, b/c if its
y ~ c1 + c2 + x1
Colinearity among c1 & c2 doesn't really hurt the slope for x, but it does often negatively affect the interpretation of the slopes for c1 & c2.
y ~ c1 + c2 + x1
Colinearity among c1 & c2 doesn't really hurt the slope for x, but it does often negatively affect the interpretation of the slopes for c1 & c2.
Every time I see these things, its alarmist, but also that post doesn't actually show the actual code and results generated so I can't really evaluate the claim.
It's always people telling me something awesome was created, w/o sharing the thing that was created
It's always people telling me something awesome was created, w/o sharing the thing that was created
October 11, 2025 at 1:55 PM
Every time I see these things, its alarmist, but also that post doesn't actually show the actual code and results generated so I can't really evaluate the claim.
It's always people telling me something awesome was created, w/o sharing the thing that was created
It's always people telling me something awesome was created, w/o sharing the thing that was created
As an analogy, I also regularly get COVID shots, and had a breakthrough COVID infection while on vacation.
That doesn't mean I'm anti-vaccine, but it still does suck that in this one specific instance, the vaccine didn't help me personally.
That doesn't mean I'm anti-vaccine, but it still does suck that in this one specific instance, the vaccine didn't help me personally.
October 10, 2025 at 3:22 PM
As an analogy, I also regularly get COVID shots, and had a breakthrough COVID infection while on vacation.
That doesn't mean I'm anti-vaccine, but it still does suck that in this one specific instance, the vaccine didn't help me personally.
That doesn't mean I'm anti-vaccine, but it still does suck that in this one specific instance, the vaccine didn't help me personally.
Yeah, I had a similar post recently and that's basically the issue.
I did a hard, time-consuming thing and learned little of interest. Doesn't mean MI is bad, but one can empathize with the disappointment.
I did a hard, time-consuming thing and learned little of interest. Doesn't mean MI is bad, but one can empathize with the disappointment.
October 10, 2025 at 3:18 PM
Yeah, I had a similar post recently and that's basically the issue.
I did a hard, time-consuming thing and learned little of interest. Doesn't mean MI is bad, but one can empathize with the disappointment.
I did a hard, time-consuming thing and learned little of interest. Doesn't mean MI is bad, but one can empathize with the disappointment.