RMD Lecture 1

RMD Lecture 1

1. measuring the strength of relationships

   • correlation

2. Bivariate correlation

   • strength?
     • verbal labels
       • 0=zero correlation .1 to .3 weak, .4 to .6 is moderate , .7 to .9 is strong
   • direction
     • positive or negative

3. correlation says nothing about causation

   • could be an intervening variable that is going on
   • effect implies causation
     • effect is not correlational

4. difficulties with correlation

   • small samples
     • ten is the minimum you can get away with
     • doesn't work with non-normal distribution: can use spearman's rho to some extent, however, if data is so far from normal the correlation coefficient that we get will not be useful
   • outliers
     • majority of cars you can buy on the high street will be comparable in speed until you put a ferrari in
   • range effect
     • you might have a relationship a particular range of data but once you move out of this range the effect doesn't hold anymore
   • hetrogenous data

5. explaining the variance

   • key idea about correlation - looking at correlation and understanding whats going on in the data
   • key idea about correlation
     • looking at correlation and understanding what's going on in the data
     • the higher the correlation, the more accurately we can predict

6. non linear relationships

   • e.g whether you have all your own teeth - low correlation, however the scatter plot does show a systematic relationship

7. partial correlation

   • a partial correlation coeffcient is usually smaller than the corresponding straight correlation efficient between intelliogence and RMD score
   • gives a clearer idea of the variance shared by two variables in the situation where other variables in a situation where other variables might also account for differences in scores

8. designing questionnaires

   • to test people
   • to test new theories
   • to understand behaviour
   • to test change
   • might use it when it is unethical

9. when creating questionnaires, must be analysed

   • we need to define concepts

10. what can we measure?

    • attributes
    • behaviours
    • beliefs
    • attitudes
    • psychometic factors

11. types of closed questions

    • likert scales
    • self-efficacy
    • semantic differentials

12. acquieent response set

    • some people will agree with everything you say
    • desirable respoonse set
    • include contradicting statements
    • include dummy and masker questions