A bias is a cause of statistical analysis errors related to the experimentation method.
Yes, we are making mistakes like everyone, but not as much, I promise ...
Attrition bias
They are due to differences between groups. They are linked to outputs test or treatment interruptions.
For example, we have 2 groups of 100 participants : group A is treated, group B received a placebo.
I know, there are numbers, but it won’t be complicated !
In group A there are 50 outputs due to treatment intolerance, and 50 improvements.
In group B, 0 output, 50 improvements, 50 stagnation.
If we do not analyse outputted patients, we have 100% improvement in Group A against 50% in group B.
However if we analyse all patients, we do not observe any difference.
I told you it wasn’t hard was it ?
Selection bias
It is due to a difference in the composition between the treated and control groups. This can influence the results of the study. Randomization is the only way to avoid selection bias.
It is related to differences in the support between the treated group and control group.
Aucun commentaire:
Enregistrer un commentaire