Statistics

Statistics

1. Hypotheses

   • Null Hypotheses
     • Something else is to blame, not what is being investigated. There is no difference between the measurements, frequencies and no relationship between the data.
     • Scientists work to demonstrate how this hypotheses is not valid.
     • First step in statistical analysis is determining what your null hypotheses is: e.g regular coffee intake has no effect on MAP.
   • They are never proven, only ever supported or rejected with data.
   • Alternate Hypotheses
     • What your'e investigating is to blame, there is a significant difference between measurements, frequencies and a relationship between data.

2. Probability Values (P Values) are generated from statistical tests.

   • P Values assist in determining which hypotheses to accept and which to reject. They test the significance of collected data.
     • 0.00 - 1.00
       • 0.00: Effect is unlikely due to chance.
       • 1.00: Effect is likely due to chance.
   • 0.05 is the generally accepted critical probability value
     • A value above (P >) 0.05: Null Hypotheses is correct and there is no significant difference. Any difference in data is likely due to chance or normal variation.
     • A value below 0.05: Null hypotheses is incorrect and Alternate is correct
     • On SPSS: P Value = .Sig

3. Normal Distribution

   • Important for quantitative (measured/numerical) data, ensures data is representative of a normal population.
     • Qualitative data: observations, narrative and descriptive data)
   • Helps determine how we present and analyse data.
   • Kolmogorov Smirov and Shapiro Wilk tests can be used to test for normal distribution
     • P Value less than (<) 0.05 means that data is not normally distributed. Present data as Median (Range) and analyse using non-parametric tests.
     • P Value more than (>) 0.05 means that data is normally distributed. Present data as a mean and use Parametric tests.
   • One of the first steps of statistical analysis is the data normally distributed or not?

4. Statistics allow us to build a model and then analyse it to draw/infer conclusions.

   • Why use statistics:
     • Use of statistics allow us to infer the meaning of collected data.
     • Good experiments consider statistics during the experiment design.
     • Science=Research=Data. Data needs analysing.

5. Decision Making Errors:

   • Type 1: Reporting a difference when there is not one.
   • Type 2: Reporting no difference when there is one.
   • Chances of making these errors are reduced by employing appropriate experimental design and analysis.
   • 50% chance of doing experiment wrong (Type 1 (alpha) and 2 (Beta) Errors.

6. Experiments:

   • Design and Hypotheses testing:
     • Ask Q's.
       • Find out what is already known.
         • Formulate a Hypotheses.
           • Test experimentally.
             • Analyse/draw conclusions
               • Support: Publish
               • Fail to Support: Publish
                 • Reformulate
   • Good Experiment:
     • One that tests for what you want to test and controls everything else.
     • Need to consider:
       • Sample Size
         • Too many and too few is bad and decreases effect of uncontrolled variation.
         • Determined by Power Calculation
           • 1 - 0.2 (Beta) = 0.8 (Means it is 0.8 SD's away from the control group.
           • Increase Power by increasing sample and effect size.
       • Control Groups:
         • No intervention
         • Needs to be concurrent
         • Beware of the Placebo effect
         • Beware Confounding Variables:

7. Terminology:

   • Treatments: Specific experimental conditions (may contain several factors)
   • Experimental Units: Subjects to treatments.