Year: 2024
Type: dissertation
Abstract: Continuous measurement results of randomized clinical trials are usually derived from a variety of measurement techniques and scales.However, to obtain a universal estimator during the meta-analysis of such trials, it is necessary to follow a common line both in the methods and in the measurement units of these trials.In cases where this is not possible, Standardized Mean Differences (SMDs) are more commonly used.SMDs contain the observed differences in means of the 2 groups (control group and experimental group) divided by their standard deviation in each trial.The advantage of SMDs in any clinical trial is that they remain independent of the way the latter is conducted, thus constituting an ideal effect size, in which the units of measurement are always expressed in standard deviations.On a practical level, however, SMDs are not easily interpretable, resulting in scientists re-using methods to draw useful statistical conclusions.These methods include splitting continuous data based on a certain threshold with the purpose of categorizing individuals into 2 groups and their subsequent comparison.In this thesis, we deal with the implementation of these methods, and more specifically with these of Hasselblad and Hedges (HH), Cox and Snell (CS), Furukawa (FU) και Suissa (SU), which use SMDs to derive other effect sizes with greater interpretability (odds ratios, risk ratios, risk differences, numbers needed to treat).For this implementation, the statistical package Stata and its functions were used.
Authors: Δεμετσιώτης, Κυριάκος Θ., Δεμετσιώτης, Κυριάκος Θ.
Language: Greek
Subfield: Statistics, Probability and Uncertainty
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7072257722