Page 1 of 1

sample size issues

Posted: Sat Jun 25, 2022 8:35 am
by taiwo
I'm working on accident involvement and severity prediction among three groups of commercial drivers using SEM(SmartPls). There is a question in the questionnaire that that states " have you being involved in an accident before?", when and how severe is the accident?" with options like no injury, no injury but car damage, slight injury,severe injury, fatal injury.
So the study involves using drivers behaviour to predict accidents involvement among commercial drivers in the country.
The questionnaire for the drivers behaviour factored into 3 latent constructs (violation, inattention errors and driving errors) with each consisting items (observed variables) of 23 in total.
The idea of the study is to compare the three groups of commercial drivers(taxi, trucks and buses) based on which have more accident involvement, which behaviours predict accident among the groups(does the groups have different distinct accident predictors?), are there different mediators across the specific groups of commercial drivers? and establish relationships between each groups based behaviours and accidents ...This summarizes the study...
The samples sizes from the 6 regions where no data seems accessible and also that I'm investigating three groups commercial driver from each regions is one major challenge.
I thought of using the G* power sample size calculator which requires number of predictors, power, etc...but after getting the sample size, i don't know if all the three groups of drivers will have equal sample size based on the G power calculator since I'm using the same number of behavioural items as predictors and same socio-demographics as mediators for the three groups of commercial drivers.