Statistical methods
No description yet. This curated deck is ready for daily study.
Card preview (7)
What do you need to know before you start your experiment?
• size of experiment • type of experiment • duration and sequencing of experiment • species, strains/stocks • adaptation • group sizes, number of groups • randomisation
Large group will affect the standard deviation how?
Larger group –> smaller SD
Why perform pilot experiments?
• Logistis - setup of experimental protocol • effect size of treatment • variation • determine size of experiment (SD important)
name shapes for dose-response curves
• linear response • nonlinear, monotonic response (slope is either positive or negative all the way) • nonlinear, nonmonotonic response (slope changes) • binary response (threshold; none at all, suddenly everything)
uni-variate statistics
• variables are • independent • large numbers studies • Data tables are long and thin can be analysed using ANOVA, t-test, linear regression, etc
today it is not uni-variate statistics, but
• large number of variables • few number of animals • data tables are short and fat This is done because • cost efficient • ethical • follow rules and regulations • other reasons needs multivariate analysis, e.g. PCA
What happens when you use classical statistics on “short and fat” data? I.e. why do we need multivariate data analysis?
• increased risk of false results (spurious results) • information is lost regarding dependent variables (co-variation)