When: Thursday 24th November 2016, 13:00-14:00.
Included in meeting:
This fortnight we have Ben and Marcela presenting at our meeting. Details on their presentations are included below.
Ben will demonstrate some code and functions useful for preparing design matrices for regressions modelling. He will demonstrate:
* dropping columns that exceed a threshold proportion of missing observations,
* dropping columns that contain little variation,
* visualising correlation matrices,
* reducing the colinearity of the design matrix by dropping columns in a manner that attempts to maximise the retained information content of the resulting design matrix.
This demonstration will use functions from the `caret` package and perform data manipulation with base R, `dplyr` and `tidyr`.
Marcela will be presenting on ‘Fast and wombled approaches to detect spatial patterns of neurodegeneration consistent with progression towards Alzheimer’s disease.’
Identification of spatial patterns are often of interest when studying underlying physical and biological processes within a neighbourhood structure. The estimation of adjacency matrix W, falls into two main categories; maximisation of a spatial statistic and estimation of W as a parameter in a Bayesian hierarchical framework known as wombling.
In this talk I’ll present both approaches and discuss its application on magnetic resonance imaging data, to identify differences in spatial patterns between healthy ageing and Alzheimer’s disease pathology.
We look forward to seeing you there!
Leah and Ace