Simulated Annealing
@simulatedannealing.bsky.social
Oncologist | RadOnc | Stats #Rstats | Opinion | Politics | Never Trump
More great stuff from @f2harrell: Optimal Bayes decisions optimize expected utility/loss/cost/objective
But different patients have different utility weights - what to do? I think run more discrete choice experiments asking patients just their priorities
But different patients have different utility weights - what to do? I think run more discrete choice experiments asking patients just their priorities
March 1, 2025 at 1:55 PM
More great stuff from @f2harrell: Optimal Bayes decisions optimize expected utility/loss/cost/objective
But different patients have different utility weights - what to do? I think run more discrete choice experiments asking patients just their priorities
But different patients have different utility weights - what to do? I think run more discrete choice experiments asking patients just their priorities
I really like Bayesian survival models with splines to estimate baseline hazard, conditional survival properties here would be highly clinically relevant, curious as to your thoughts on advantages here vs Frequentist framework @f2harrell.bsky.social
arxiv.org/pdf/2002.09633
arxiv.org/pdf/2002.09633
February 21, 2025 at 7:07 PM
I really like Bayesian survival models with splines to estimate baseline hazard, conditional survival properties here would be highly clinically relevant, curious as to your thoughts on advantages here vs Frequentist framework @f2harrell.bsky.social
arxiv.org/pdf/2002.09633
arxiv.org/pdf/2002.09633
I love this table here; “hopefully” is key even with good artificial control, more often P1 !=P2, and causal inference cannot rescue
February 18, 2025 at 4:26 PM
I love this table here; “hopefully” is key even with good artificial control, more often P1 !=P2, and causal inference cannot rescue
‼️ WOLVERINE
IPD meta-analysis of 5 omPC RCTs of MDT and standard of care (SOC) vs. SOC alone
OS HR 0.63 (0.39-1.0)
IPD meta-analysis of 5 omPC RCTs of MDT and standard of care (SOC) vs. SOC alone
OS HR 0.63 (0.39-1.0)
February 15, 2025 at 10:10 PM
‼️ WOLVERINE
IPD meta-analysis of 5 omPC RCTs of MDT and standard of care (SOC) vs. SOC alone
OS HR 0.63 (0.39-1.0)
IPD meta-analysis of 5 omPC RCTs of MDT and standard of care (SOC) vs. SOC alone
OS HR 0.63 (0.39-1.0)
To illustrate this point, look at how non-quadratic the loglikelihood plot is for example w/ binomial logistic regression - very commonly used!
How many Frequentist CIs in the literature are wrong?
How many Frequentist CIs in the literature are wrong?
February 9, 2025 at 2:56 PM
To illustrate this point, look at how non-quadratic the loglikelihood plot is for example w/ binomial logistic regression - very commonly used!
How many Frequentist CIs in the literature are wrong?
How many Frequentist CIs in the literature are wrong?
Saturday night thought: assuming linearity in continuous predictors is widespread and potentially disastrous. Checking Martingale residuals plots is essential! HT @f2harrell.bsky.social Ch. 20 #rms
January 18, 2025 at 11:51 PM
Saturday night thought: assuming linearity in continuous predictors is widespread and potentially disastrous. Checking Martingale residuals plots is essential! HT @f2harrell.bsky.social Ch. 20 #rms
❗️ RCT of proton RT vs TACE in HCC
mPFS for PBT vs TACE was not reached vs. 12 months, p = .002; with a whopping HR 3.62 (1.62–8.05)
TACE is a very poor treatment 👎
mPFS for PBT vs TACE was not reached vs. 12 months, p = .002; with a whopping HR 3.62 (1.62–8.05)
TACE is a very poor treatment 👎
January 16, 2025 at 11:53 PM
❗️ RCT of proton RT vs TACE in HCC
mPFS for PBT vs TACE was not reached vs. 12 months, p = .002; with a whopping HR 3.62 (1.62–8.05)
TACE is a very poor treatment 👎
mPFS for PBT vs TACE was not reached vs. 12 months, p = .002; with a whopping HR 3.62 (1.62–8.05)
TACE is a very poor treatment 👎
High yield workflow right here for checking OLS assumptions. Thank you Frank for more great work
January 16, 2025 at 1:09 PM
High yield workflow right here for checking OLS assumptions. Thank you Frank for more great work
TOPGEAR
Ph 3, GEJ/gastric cancer -> periop chemo +/- preop CRT -> surgery
CRT benefit by age and primary tumor site? 🧐
Ph 3, GEJ/gastric cancer -> periop chemo +/- preop CRT -> surgery
CRT benefit by age and primary tumor site? 🧐
January 12, 2025 at 2:17 PM
TOPGEAR
Ph 3, GEJ/gastric cancer -> periop chemo +/- preop CRT -> surgery
CRT benefit by age and primary tumor site? 🧐
Ph 3, GEJ/gastric cancer -> periop chemo +/- preop CRT -> surgery
CRT benefit by age and primary tumor site? 🧐
January 11, 2025 at 9:51 PM
January 11, 2025 at 9:48 PM
January 10, 2025 at 11:17 PM
As bad as it gets in terms of misinterpretation? HR 0.82 with 0.95 CI (0.63-1.05) = very likely there is benefit, NOT “no benefit” 🤦
January 10, 2025 at 4:01 AM
As bad as it gets in terms of misinterpretation? HR 0.82 with 0.95 CI (0.63-1.05) = very likely there is benefit, NOT “no benefit” 🤦
‼️ EUROPA
Stage I HR+ breast cancer in women 70+ -> lumpectomy -> APBI vs endocrine
Better QoL with APBI! Very important trial!
#radonc @icromeattini.bsky.social
Stage I HR+ breast cancer in women 70+ -> lumpectomy -> APBI vs endocrine
Better QoL with APBI! Very important trial!
#radonc @icromeattini.bsky.social
January 10, 2025 at 2:55 AM
‼️ EUROPA
Stage I HR+ breast cancer in women 70+ -> lumpectomy -> APBI vs endocrine
Better QoL with APBI! Very important trial!
#radonc @icromeattini.bsky.social
Stage I HR+ breast cancer in women 70+ -> lumpectomy -> APBI vs endocrine
Better QoL with APBI! Very important trial!
#radonc @icromeattini.bsky.social
SUPREMO
Mastectomy + ALND +/- PMRT in “intermediate risk” breast cancer
pN1 OS HR 0.82 (0.63-1.05)
“No benefit” 🤦 = misinterpretation of the main results of their own work which is consistent with 18% reduction in death and can only rule out > 37% reduction #radonc
Mastectomy + ALND +/- PMRT in “intermediate risk” breast cancer
pN1 OS HR 0.82 (0.63-1.05)
“No benefit” 🤦 = misinterpretation of the main results of their own work which is consistent with 18% reduction in death and can only rule out > 37% reduction #radonc
January 10, 2025 at 2:17 AM
SUPREMO
Mastectomy + ALND +/- PMRT in “intermediate risk” breast cancer
pN1 OS HR 0.82 (0.63-1.05)
“No benefit” 🤦 = misinterpretation of the main results of their own work which is consistent with 18% reduction in death and can only rule out > 37% reduction #radonc
Mastectomy + ALND +/- PMRT in “intermediate risk” breast cancer
pN1 OS HR 0.82 (0.63-1.05)
“No benefit” 🤦 = misinterpretation of the main results of their own work which is consistent with 18% reduction in death and can only rule out > 37% reduction #radonc