At the start of October I started my PhD on patient allocation methods within clinical trials. Randomised controlled trials are at present the approach most often used in clinical research, however they do not give any possibility to change the treatment allocation probability within the trial.
This week all the STOR-i MRes had to pick a topic within statistics or operational research to write a longer literature review on. I chose to write about how mathematical models can be used to track the spread of diseases, such as influenza.
Bayesian statisticians interpret probability as the subjective experience of uncertainty. Bayes’ theorem is a model for learning from data. There is no notion of infinitely repeating an event of interest. Rather you use prior information to estimate an event outcome. Once this outcome is revealed, then prior information is updated. This is the model of learning from experience (data).
Within statistics we often have a known stochastic model for collected data. This model will frequently include unknown parameters, which we want to make inference about. Estimates of the data parameters can be found using likelihood inference.
Optimisation is the method used to either maximise or minimise an objective function, by picking particular values of variables. These variables must follow a number of constraints. Non-linear programming is optimising an objective function, where the objective function and all constraints are non-linear.
Optimisation is the method used to either maximise or minimise an objective function, by picking particular values of variables. These variables must follow a number of constraints. Linear programming is optimising an objective function, where the objective function and all constraints are linear.
I now describe a specific design of a clinical trial using a multi-armed bandit model. This model uses a Bayesian adaptive design to choose an allocation method which maximises the total expected reward.
Last week all the MRes students at STOR-i were asked to write a short literature review on a part of statistics or operational research that we found interesting. I chose to write about how multi-armed bandits could be used when designing a clinical trial.
Last week I was speaking to a STOR-i alumni Helen Barett, who told me about her PhD project in pharmacokinetics. Part of her PhD was about how microsampling could be used in pharmacokinetics and toxicokinetics and how it compared to the traditional approaches of sampling.
This week I was speaking to a STOR-i alumni Helen Barnett. She spoke to me about her PhD project in pharmacokinetics, which I found really interesting. Her project reminded me of my fourth year dissertation titled, 'Chemokine gradient development in the lymphatic interstitium', which I completed last summer. This investigates how chemokines cause white blood cells (leukocytes) to move around the body.