Workshop: Experimental Design & Health Statistics

 

Experimental Design

-          missing data and out of range data à need to know what to do

-          need to control lurking variables

 

-

-          Different definition used of survey from that used in part 1

o   survey versus experiment  à ie survey = observational

o   in de Vaus, a survey is a data collection method

-          Experimental units

o   CSU – experimental unit and observational unit

§  can be one and the same

§  can be different

§  experimental unit – physical unit to which treatment applied

§  observational unit – basic unit on which a measurement is made

-          Need to clarify à what is replication?

 

"the repetition of the set of all the treatment combinations to be compared in an experiment. Each of the repetitions is called a replicate."

 

o   is the power of 2 replicates the same as one experiment with same sample size as the combined replicates

o   what is the relationship between a replicate and a meta analysis

 

 

-          randomization à same probability of being selected for treatment

-          need to understand definitions of basic heath research designs

-          randomness / not of surnames

o   some cultural groups could have names beginning with certain letters

o   see article of questionnaire design

o   but within Anglosaxon Australia, first letter of name could be random??

 

 

-          Need to be careful when deciding if experiment has been blocked

 

 

Health Studies

 

-          lifecycle

-          birth à  life time  à death

-          fertility à mobidity à mortality

 

 

-          fertility / demography

-          mobidity à epidemiology

 

 

 

-          look at calculations

o   where people lost to followup

 

 

-          what is impact of death on these calculations

 

 

-          relate incidence and prevalence to Baysian issues

 

-          Crude death rate à does not allow for age mix  / age structure of different populations

-          age specific death rate

-          standardization of rates

 

 

 

-          randomized trial à how are drop outs handled

 

-          measure of risk

o   relative risk / odds ratio

o   do these measures have confidence intervals

-          why not use chi squre

 

 

-          odds ratio à compares occurring versus not occurring

-          clarify why we cannot calculate risk ratio from case control study