Inferential statistics

• what are inferential statistics?

  Inferential statistics tell us wether to accept or reject the null hypothesis by using probability
• Criteria for using a parametric test

  No need to calculate but I need to know the 3 criteria

  e.g. What are the criteria for a parametric test (3 marks)

  • SD is not significantly different (scores are very near the mean score)
  • Data is normally distributed
  • Data needs to be interval or ratio
• Criteria for non-parametric test (x5)

  • the type of prediction made (difference or correlation between two things )
  • the type of experimental design used (repeated measures or independent measures)
  • the level of data obtained (nominal, ordinal or interval)
• What are the 5 non-parametric tests

  • chi squared
  • binomial sign test
  • Mann-Whitney U test
  • wilcoxon T test
  • Spearman’s rho
• Why do we use inferential stats instead of descriptive statistics (measures of central tendency or measures of dispersion)?

  inferential stats tell us wether the results are simply down to chance or due to manipulating the IV whereas those measures don’t
• p<0.05

  results being down to chance is LESS than 5% due to chance

  results are SIGNIFICANT
• p>0.05

  results being down to chance is MORE than 5%

  results are NOT significant
• what is a type 1 error?

  False positive: we falsely accept the alternative hypothesis and reject the null
• what is a type 2 error?

  false negative: we falsely accept the null hypothesis and reject the alternative
• What is the rule of R for chi square and spearman’s rho?

  Calculated/observed value must be equal to or greater than critical value
• what is the rule (of R) for binomial sign test, Mann whitney U and Wilcoxon T test?

  calculated/observed value must be equal to or less than critical value
• Decisions for non-parametric tests 


  .

  
• Binomial sign test

  1. Calculate the difference between two conditions (a-b) record + for positive difference and - for negative difference, ignore 0 difference and remove from total P number
  2. Add up the number of + and - signs, whichever is the smallest is S

  
• Mann Whitney U test


  1. one condition with fewer P’s (condition A)
  2. Rank scores as if they belong to one group
  3. Sum of the ranks (R1 and R2)
  4. U1 and U2
  5. U is the smaller of U1 and U2
  6. U must be equal to or less than the critical value to be significant

  
• Chi squared

  O= observed value

  E= expected value

  Criteria: testing for a difference, IMD and nominal data

  
• Spearman’s rho

  
  

  1. Rank values of first and second variable SEPARATELY
  2. find the difference between ranks
  3. Square the difference
  4. Add up squares and insert into equation for rs
  5. IGNORE - signs and KEEP 0 scores

  
• Wilcoxon T test

  .

  1. Work out the difference between the scores in the two conditions
  2. Rank the differences form the smallest to largest, ignore +/- signs and if there is a 0 difference REMOVE from total participant number, T is the smallest of the values

  
• For binomial sign test and Wilcoxon T test, you IGNORE a 0 difference and remove from the total no. of participants
• All non parametric tests test for a difference except…

  Spearman’s rho
• what does spearman’s rho test for?

  a correlation
• what non-parametric test do you rank values as if they belong to one group?

  Mann Whitney U test