- Видео 8
- Просмотров 28 511
Alan Maloney
Добавлен 5 фев 2013
There are two target audiences for these tutorials. The "Pharmacometric Tutorials - General Principles" would be suitable for anyone working within clinical drug development, especially those involved in the design, analysis and interpretation of phase 2 and phase 3 studies. The "Pharmacometric Tutorials - Technical Issues" is aimed at pharmacometricians and biostatisticians involved with Modelling and Simulation, PK/PD analysis etc., working towards a Model Based Drug Development approach to clinical drug development. The wonderful "Khan Academy" story and Lewis Sheiner helped motivate the production of these videos.
Hamiltonian Monte Carlo For Dummies (Statisticians / Pharmacometricians / All)
Hamiltonian Monte Carlo (HMC) is the best MCMC method for complex, high dimensional, Bayesian modelling. This tutorial aims to provide an introduction to HMC through worked examples ranging from elementary to complex models.
Просмотров: 13 676
Видео
PAGE Conference Presentation - June 2017
Просмотров 3817 лет назад
This is the presentation for the PAGE (Population Approach Group in Europe) in Budapest, June 2017 entitled "The 6 Biggest Pharmacometrics Modelling Mistakes"
L6 - Statistical Modelling - Uncertainty and Sensitivity Analysis
Просмотров 11 тыс.10 лет назад
All predictions and simulations from statistical models require an understanding of how to reflect uncertainty and utilise sensitivity analysis. This tutorial explains these concepts and highlights key points to consider.
L5 - Individual and Population Dose Response
Просмотров 55110 лет назад
This tutorial explains the difference between individual and population dose response, and the consequence for drug development and regulatory approval.
L4 - Phase 2 Study Design - Technical Details
Просмотров 18210 лет назад
This is a technical tutorial covering the main components of phase 2 study design. The primary audience will be pharmacometricians and biostatisticians who are involved in the design and analysis of phase 2 dose response studies.
L3 - Why the Sigmoidal Emax model is Special
Просмотров 2 тыс.10 лет назад
The best model for describing a dose response relationship if often the Sigmoidal Emax Model. This tutorial explains why.
L2 - Phase 2 Study Design - General Principles
Просмотров 38010 лет назад
This tutorial discussed the design of phase 2 studies, and is suitable for those engaged in the design, conduct, and analysis of clinical drug studies.
L1 - The Goal of Cinical Drug Development
Просмотров 22110 лет назад
An introductory tutorial explaining the goal of clinical drug development within a model based drug development (MBDD) framework.
Thank you for great explanation ❤
Thanks for the amazing video. It helped a lot!!
Jeez after watching this and Ben's Lambert, I am ready to try by my own the HMC, thanks a lot that sucha good presentation on the intuition of the HMC.
Glad you found it useful!
Hi, this is the same three-parameter Hill model wich are cited on Toxcast EPA project?
I expect so, but just compare the two equations to be sure.
very confusing
Thanks for the insightful video. Could you please provide some sample codes of the figures you show on the slides? Thank you
Thank goodness for RUclips recommendation.
Sorry may be I missed it but did you discuss how it can traverse other contours? Your example showed how it can stay in its contour plane.
Hi Y. Good question. HMC follows the joint contour 'within' an iteration (i.e. for the L=15 steps) ...see the video at around 9:48, where it followed the contour with m=1.00 for iteration 1 to yield the new value for theta (e.g. -0.65 in the video). For the next iteration, it will sample a new m variable (say m=0.00). This will 'shift' us to a new joint contour for iteration 2 (the intersection of theta = -0.65 and m=0.00 on the graph). Thus 'between' iterations is when we "jump" between the joint contours. Hope this is clear.
I have some doubts. What does stan do that other languages like R or python doesn't? Can I use stan for constrained optimisation? I have a problem I need help with. Its about maximising an objective function subject to set of inequality constraints. Would you be abe to help or give some directions?
Hello. R currently doesn't have anything like Stan. I am not sure about Python, but I doubt it. Stan is not built to solve constrained optimisation problems - rather to fit complex Bayesian models. My only advice would be to google for tools/packages to solve constrained optimisation problems like yours. For simple problems, I expect many tools would have something. However for more complex problems, you may need more specialised software. Good luck. Al
@@alanmaloney2791 My maximisation function is a log likelihood function, similar to how we solve MLE, but subject to some constraints regarding a distribution parameter theta. So not sure if it counts as a "Bayesian".
Hello. If the constraint is on a parameter, Stan can handle that (see their website). If you have many constraints like theta1+theta2+theta3 = 100, theta1*theta3 = 10 etc., I think it may be more tricky to code in Stan. If you are wishing to do MLE, other software may be easier. Cheers Al
Great blog. The best HMC video I have seen before. Thanks.
Thanks for the friendly feedback...delighted you liked it!
Thanks Alan. So inisgthful and well explained
Thanks Quim for the friendly and positive comment. Much appreciated!
Dear Alan, I work on clinical trials and I found your video immensely helpful for me to understand the exact question on my current program. I enjoyed your presentation and I hope more pharmaceutical researchers will come here and benefit from your video. Thank you so much! Please consider making more videos on these interesting statistical topics!
Hi Wenyan, Many thanks for your friendly note. This presentation was partly a forerunner to a paper that was published in CPT (ascpt.onlinelibrary.wiley.com/doi/abs/10.1002/cpt.710); I am trying to get pharma/regulators to think in terms of maximising individual patient outcomes via optimal dose ranges (not "one size fits all" dosing). A copy of the paper can be found on the "About" page of my diabetes website (www.comparediabetesdrugs.com/about/)...under the first "read more" section. Any comments welcome! Thanks again, Al
Video aside, I don't think any real dummy is interested in HMC, lol.
Thank you for this very illustrative and insightful video.
Great to hear - delighted someone thought it was useful!
I want to perform global sensitivity and uncertainty analysis considering pdfs ..how to do this please..can anyone help me ?
Bonjour, Alan Maloney. this is a colorful video. thanks. :)
This is interesting. Are there any published papers treating the issue of the fixed Hill coefficient?
Hi...I am not aware of any, but have not really looked. It is also worth mentioning that if you use a) dose or b) a measure of average drug exposure (e.g. average concentration at steady state (Css)) or c) use actual/predicted concentrations over time to drive the pharmacodynamic response, then Hill cannot be simultaneously equal to 1 under all three. Hence if you set it to 1 using dose, then you are implicitly setting it to something different to 1 based on concentrations. In short, do not fix it!! Estimate it, just like we would an ED50 (or EC50). cheers Al
Thank you Alan for this video.
Great video. I was wondering when you discussed the problems of the richards model you discussed the difficulty of getting enough data even with the emax. I was wondering, is there a rule of thumb for the number of observations for the Emax? Also, is there some seminal paper about Emax or dose response in experimental context? I am in economics so I might not be up there in that particular area... I already have some paper on dose ranging with Emax... Anyways very useful! Thanks!
Delighted you thought the video was OK. Alas there is no rule of thumb for the number of observations, as it is always a combination of the size of the residual error variance (our noise), the design (i.e. what doses you picked), and the precision you want (wide or narrow uncertainty in the final estimated dose response). You can pick the doses in an optimal way (see L4 lecture), based on what you want to optimise. Maths wise, we 'know' the asymptotic behaviour of the variance-covariance matrix is the inverse of the expected Fisher Information Matrix. If you want the formulae, just email me at the address in the video. If maths is not your strong point, just simulate your design and estimate it (say 20-1000 times). You can then play with different designs (doses and observations per dose) to see how they do relative to your metric(s) of interest. As a rule of thumb, D optimal designs for this model are 4 point designs, at doses that yield 0%, 26%, 74% and 100% of the response (although we must replace the 100% dose with our highest dose). Good luck.
Hey mate, thanks for the video. What software are you using?
Hi Ken...this was done using SAS 9.4
Thank you so much. It is very helpful!
Sir Thank you very much for making this video. I am working on MICROWAVE water cloud model sensitivity analysis. Below i am describing its equation and its parameter, WCM=A*v1*cosd(40)*(1- exp(- 2*B*v2*secd(40))) + exp( - 2*B*v2* secd(40))*(C + D*MV) x <- c(.11, .169, .235, .235, .301, .432, .432, .999, .999, 1.031, 1.063, 1.158, 1.318, 1.241, 1.088, 0.973, 1.359, 1.359, 1.444, 1.784, 1.774 ,1.58, 1.58, 0.806 ,0.896 , 0.986, .986, 1.077, 1.258, 1.509, 1.262) y <- c(-16.71, -20.16, -17.54, -16.52, -18.79, -18.34, -18.64, -15.91, -15.15, -14.8, -12.87, -16.2, -15.03, -15.02, -14.7, -16.88, -14.87, -12.8, -12.48, -10.97, -9.51, -12.34, -12.27, -11.93, -9.14, -10.46, -8.6, -8.86, -10.29, -9.88, -10.53 ) relation <- lm(y~x) print(relation) Call: lm(formula = y ~ x) Coefficients: (Intercept) x -18.550 4.671 Where WCM = water cloud model in WCM equation parameter C= intercept -18.550 and D= x 4.671 A is contant lies between 0.17 to 0.25 B is contant lies between 0.12 to 0.19 v1=v2= 0.20 I am unable find its sensitivity and uncertainty by R programming .I am very inspire by your you tube video. Kind request to you please help me its sensitivity and uncertainty analysis. I am waiting your positive reply
Hello...glad you liked the video. Alas I am no expert in R, but I would guess the LM function would be providing the parameter uncertainty...check out the documentation for it online...good luck!
Sir Thank you for rely. I am searching but not get. There is a package(sensitivity) for this but i am unable to execute this. If you help me then i am grateful of you.