Implementation, review and development of methodology for single case comparisons to small samples in the R package singcar
(2021) STAN40 20211Department of Statistics
- Abstract
- Statistical comparison of single cases to small samples is a methodology that has
been extensively used in, for example, cognitive and clinical neuropsychology. This
is most often done to determine changes in cognitive processing after an individual
has incurred some type of brain damage. In a clinical setting one often wish to
infer whether a patient exhibit abnormally low performance on some cognitive
ability. In cognitive neuropsychology on the other hand one often wish to infer if
a patient exhibits an abnormally large discrepancy in performance between two
cognitive abilities. Because cognitive abilities seldom are well represented by one
single performance on one single task one might additionally be interested in the
... (More) - Statistical comparison of single cases to small samples is a methodology that has
been extensively used in, for example, cognitive and clinical neuropsychology. This
is most often done to determine changes in cognitive processing after an individual
has incurred some type of brain damage. In a clinical setting one often wish to
infer whether a patient exhibit abnormally low performance on some cognitive
ability. In cognitive neuropsychology on the other hand one often wish to infer if
a patient exhibits an abnormally large discrepancy in performance between two
cognitive abilities. Because cognitive abilities seldom are well represented by one
single performance on one single task one might additionally be interested in the
abnormality of a case on several measurements converging on a cognitive ability of
interest, or the abnormality of a case in a multivariate space. Several methods to
estimate case abnormality have been developed that keeps the Type I error rate at
its nominal level. However, they have not been available in any standard statistical
software environment and their documentation is spread thin across multiple articles
and compiled computer programs. This thesis aims to gather and review the most
popular methods while presenting them and their usage in the R package singcar.
Of note are the more flexible and useful methods that have not received as much
spread as the simpler. These include techniques using Bayesian regression to allow
for the inclusion of covariates and linear mixed models to handle repeated measures
data. Additionally, statistical comparison of single cases to a control population
are inherently low powered. To facilitate experimental planning and design power
calculators have been implemented in singcar and the concept of power for this
type of statistical analysis is reviewed. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9049385
- author
- Rittmo, Jonathan LU
- supervisor
- organization
- course
- STAN40 20211
- year
- 2021
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- Single case research, R, Outlier detection, Patient diagnostics, Package development
- language
- English
- id
- 9049385
- date added to LUP
- 2021-10-20 08:52:27
- date last changed
- 2021-10-20 08:52:27
@misc{9049385, abstract = {{Statistical comparison of single cases to small samples is a methodology that has been extensively used in, for example, cognitive and clinical neuropsychology. This is most often done to determine changes in cognitive processing after an individual has incurred some type of brain damage. In a clinical setting one often wish to infer whether a patient exhibit abnormally low performance on some cognitive ability. In cognitive neuropsychology on the other hand one often wish to infer if a patient exhibits an abnormally large discrepancy in performance between two cognitive abilities. Because cognitive abilities seldom are well represented by one single performance on one single task one might additionally be interested in the abnormality of a case on several measurements converging on a cognitive ability of interest, or the abnormality of a case in a multivariate space. Several methods to estimate case abnormality have been developed that keeps the Type I error rate at its nominal level. However, they have not been available in any standard statistical software environment and their documentation is spread thin across multiple articles and compiled computer programs. This thesis aims to gather and review the most popular methods while presenting them and their usage in the R package singcar. Of note are the more flexible and useful methods that have not received as much spread as the simpler. These include techniques using Bayesian regression to allow for the inclusion of covariates and linear mixed models to handle repeated measures data. Additionally, statistical comparison of single cases to a control population are inherently low powered. To facilitate experimental planning and design power calculators have been implemented in singcar and the concept of power for this type of statistical analysis is reviewed.}}, author = {{Rittmo, Jonathan}}, language = {{eng}}, note = {{Student Paper}}, title = {{Implementation, review and development of methodology for single case comparisons to small samples in the R package singcar}}, year = {{2021}}, }