Experimental Design

• independent data points


  data values of an individual which are unaffected by data values of another individual
• issue with pseudo-replication


  statistical tests generally assume data point is independent, will give you false positive/negative results
• What considerations, if any, should limit the number of additional “explanatory variables” you collect in practice?


  Practicality of measuring/mitigating/removing explanatory variable. Remove if possible, measure if not. Money/time/resources might prevent limiting explanatory variables                                                               
• Minimise random variation


  lab engineered organisms, same species/population/sex/age/genetics
• disadvantages of removing all random variation

  natural biology full of random variation, removal reduces application of results in real world, results not representative of oopulation
• relationship between amount of random variation in data set and magnitude of residual sum of squares (RSS)


  larger RSS means increased amount of random variation, in relation to explained sum of squares
• what does the size of RSS relative to ESS tell you how well your model explains data


  larger RSS and smaller ESS, model does not explain data well
• hypothesis


  clearly stated postulated description of how an experimental system works
• null hypothesis


  no relation between explanatory and response variables
• alternative hypothesis 


  relation between explanatory and response variables, can be directional
• null hypothesis rejected and alternative accepted


  p<0.05
• manipulative study


  experimenter alters aspect of experimental system then studies effect of change
• correlational study


  no alter in experimental system, makes use of naturally occurring variation to look at effect of one factor on another
• disadvantages of manipulative study

  unethical, doesnt represent population, doesnt take into account natural variants, not always possible/practical
• advantages of manipulative study

  takes into account random variables, easy to observe
• disadvantages of correlation study


  confounding variables, ethics, endangered species, tarnish environment
• advantages of correlational study


  natural variation, realistic, representative of population, can look at longevity, cheaper, can be more practical
• why are you at greater risk of drawing the wrong conclusions about your hypothesis if your study uses a small sample size


  results may be as a result of random variation, outcome may be down to chance, outlier can have greater effect on mean
• effect of large sample size on stats

  more measurements and calculate mean removes random variation, less likely to incorrectly accept/reject null hypothesis
• haphazard selection


  no systematic method or randomisation, easily accessible/convenient
• self-selection


  individuals volunteer to participate, introduces bias
• random selection


  individuals chosen entirely by chance, each member of population has equal opportunity to be selected
• pseudoreplication


  replicates are not independent of oneanother
• confounding variables


  third-variables affect dependent variable but not part of model
• cohort effect


  outcomes of study influenced by specific characteristics of data pool
• why does precision of estimate improve as sample size increases

  cancels some of the random variation and decreases standard error
• what limits the total sample size

  money, time, availability, resources
• false positive

  type I error
• type I error


  explanatory variable has no effect on response, but statistical test mistakingly suggests there is
• type II error


  statistical test fails to detect the explanatory variable has an effect on the response
• false negative


  type II error
• power


  probability that a particular experiment will detect an effect, probabiliyu of not making a type II error
• longitudinal study


  data collected with from same subjects repeatedly over a period of time
• cross-sectional studies


  capture data at a single point in time
• advantages of longitudinal study

  observe how variables evolve, determine patterns of change, assess cause-and-effect relationships over time
• within subjects


  experimental subjects experiences different experimental treatments sequentially; comparisons being made on the same individual at different time
• advantages of within subjects


  less variation, removes confounding variables, cleaner results, removes random variation, requires smaller sample size
• why is within-subjects not pseudoreplication


  different response and explanatory variable measures
• within subjects disadvantages 


  period effects, carry over effects, time consuming, ethical considerations
• period effects


  time could be confounding factor