Probability & significance

• Descriptive statistics may show a difference in your results; they describe & summarise the data. To check the difference (or association if doing a correlation) is significant, you need to do a statistical test.
• Inferential statistical tests allows us to make this judgement & in psychology, the accepted level of probability is 5%, eg there is a 5% risk or less that our results have occurred by chance- so we can say with 95% confidence that our results are down to the manipulation of the IV.
• Usually expressed as p<0.05 & if question doesn't state %, always assume it's 5%. Inferential statistics allow you to infer the power or significance of the data.
• A null hypothesis- states there will not be a significant difference/ correlation, should be trying to falsify our hypothesis for good science.
• An alternative hypothesis- states that there will be a significant difference/ correlation.
• Only one hypothesis can be true, accept one & reject the other.
• Type 1 error:

  • False positive
  • Saying it was significant when it actually wasn't.
  • Rejecting the null hypothesis when it's actually true.
  • When we do inferential tests and get a result below our level of significance, we accept alternative hypothesis & reject the null hypothesis.
  • Significance level may be too high, so we have concluded that there's a significant relationship/ correlation, but there really isn't.
• Type 2 error:

  • False negative
  • Saying it wasn't significant, when it actually was.
  • Accepting the null hypothesis when it's actually false.
  • When we do a test and get a result above our level of significance, we accept the null hypothesis & reject the alternative hypothesis.
  • Significance level may be too low, so we have concluded that there's not a significant relationship/ correlation, but there really is.
• The significance level is set at P<0.05 to try and reduce either of these errors happening- seen as a balanced medium.