STAT1201 WEEK 1 LEARN X · when you see variability in data, it can be hard to tell if it is due to natural variability or to experimental error ERROR VARIABILITY · can occur from measurement inconsistency, this can be reduced by giving a protocol for exactly how a measurement should be taken · this presence of error also makes it necessary to replicate experiments Group Variability · allows you to separate the data into various groups (male/female), infers consistency between individual groups and their difference to other groups Sampling Variability · allows you to use the averages in the study to say something about people in general · however if you conducted the study again you may get different data LECTURES VARIABLES Quantitative - discrete / continuous Discrete · can only have whole numbers (0, 1, 24, 45) · e.g. number of phone calls, age Continuous · is measurable, can have any value over some range · e.g. height, weight Categorical - nominal / ordinal Nominal · categories or groups that do not have an order
· e.g. marital status, country of birth Ordinal · categories or groups that have an order · e.g. academic grades, high/medium/low OBSERVATIONAL AND EXPERIMENTAL STUDIES Observational · researcher observes part of population and measures the characteristics of interest · make conclusions based on the observation but does not influence to change existing conditions or does not try to affect them · e.g. examining the effect of smoking on lung cancer Experimental · researcher assigns subjects to groups and apply treatment(s) to group(s) and the other group does not receive the treatment · can be designed as a blind or double blind study (when neither researchers/administrators of treatment, or patients know which group is what) · when an experiment involves both comparison and randomisation then we call it a randomised comparative experiment · e.g. effect of caffeinated drinks on blood pressure HYPOTHESIS TESTING Null Hypothesis (H0) · usually statement of "no effect" Alternative Hypothesis · usually statement of "an effect" · also refers challenges to status quo (something new is occurring compared to past) CONCEPT OF P-VALUE Strong evidence against null hypothesis - (0 - 0.01) Moderate evidence against null hypothesis - (0.01 - 0.05) Weak evidence against null hypothesis - (0.05 - 0.1) No evidence against null hypothesis - (0.1 - 1) WEEK 2
LEARN X VISUALISING CONTINUOUS DATA Dot Plot · good for showing details and comparing smaller groups, less useful for larger groups · when lots of people have same height dots end up being printed over and over again, hence doesn't provide an insight into pattern or distribution · one way to improve this is 'alpha blending' giving each point some transparency shows where points are overlapping · you can also spread the data into groups (male/female) to group a distribution difference Histogram · points divided into several 'bins', the more bins you have the more detailed analysis of the data you will have · useful tool to illustrate density