Research manifesto

CERG Home Page  

1. Standard statistical techniques have assumptions that are critical in interpreting the output from data analysis
   • Normality of the underlying population
   • Homogeneity of the variability in samples from such populations
   • Residuals having a normal distribution

   • Etc.

   Research often relies on having large numbers so that the assumptions become less of a problem

2. Conventional statistical hypothesis need not be very informative .

   Most statistics test all or nothing hypotheses:

   Males perform poorer than females on programming assessment tasks

   These hypotheses cannot give an indication of magnitude of effect.

   The difference between males and females on programming assessment tasks is greater than 5 assessment units with females scoring higher.

   The conventional hypothesis does not allow us to progressively test structured conceptual developments in theory construction

   If we want systematic development of understanding we have to be able to systematically test hypotheses.
   For example, we want to be able to test hypothesis about different levels of performance between males and females. We want to be able to look for the critical sources of differences. Under conventional approaches we test whether there is or is not a difference.