Inferential Statistics

• Descriptive statistics describe patterns and trends in a data set.

  They may also describe averages:

  • measures of central tendency (mean, mode, median)
  • measures of dispersion (range, standard deviation)
• Inferential statistics are used to make inferences/predictions about a whole population based on results from a sample.

  • Evaluation: GENERALISABILITY
• What are inferential statistics based upon 


  The probability that a particular outcome has arisen by chance or not.
• Findings due to chance = not significant

  Findings not due to chance = significant
• Null hypothesis 


  There will be no difference...
• Alternate (one tailed/directional hypothesis)


  predicts the direction of the relationship (e.g. bigger, smaller, more, less) of the variables
• Alternate (two tailed/non directional hypothesis)


  doesn't predict the direction, but rather states if it is thought there will or won't be a difference
• Which hypothesis is rejected/which is accepted if the results are said to be significant 


  • null hypothesis rejected
  • alternate (d/nd) accepted
• Which hypothesis is rejected/which is accepted if results are said to be not significant 


  • alternate hypothesis rejected
  • null hypothesis accepted
• EXAM TIP: A HYPOTHESIS IS ALWAYS NON DIRECTIONAL UNLESS THE QUESTION STATES THERE HAS BEEN PREVIOUS RESEARCH CONDUCTED
• What is a type 1 error

  Type 1 error = false positive

  • Occurs if the researcher REJECTS THE NULL which was actually the true hypothesis
  • (says the results are significant when they are not)

  Too lenient - significance level is too low
• What is a type 2 error

  Type 2 error = false negative

  • Occurs if the researcher ACCEPTS THE NUL which was actually the incorrect hypothesis
  • (says the results are not significant when they are)

  Too strict - significance level is too low
• Why do psychologists use a certainty of 95%


  In a scientific study you can't be 100% sure whether data has occurred due to chance or if it occurred due to a real difference/correlation between trials

  • Therefore psychologists say they are 95% certain that the results obtained are not due to chance or unknown errors
• The 5% level is recorded as p=0.05/p<0.05

  • the level of acceptance is the significance level
• In some studies psychologists have to be more certain such as in drug trials or if they want to challenge the significance of a previously conducted study.

  • They would therefore increase the significance level to 1%
  • This means p=0.01 and psychologists say that there is only a 1% chance that results are due to chance and not real difference/correlation.
• Where p=0.05 for a one tailed/two tailed test and the value of n is ... The calculated value of S/T/U... is greater/equal/less than the critical value which indicates the results are/aren't significant.
• How do we calculate n 


  N is the number of PPT's
• What is a test of difference


  experimental - 'there will be a difference'
• What is a test of association 


  correlation - 'there will be a relationship'
• What types of data are used in parametric tests

  * parametric tests are the most powerful statistical tests, using measures of dispersion.

  • They use only interval data
• What types of data are used in non-parametric tests 


  ranked data (loses detail)

  • nominal/ordinal
• What is nominal data 


  Objective, categorised data

  • e.g. boys and girls
• What is ordinal data

  data that is ranked (it isn't fixed or objective)

  • e.g. happiness from most to least
• What is interval data 


  objective data given along a fixed point scale

  (e.g. temperature)
• What is related data 


  matched pairs/repeated measures - the PPT repeats both conditions
• What is unrelated data 


  independent groups - the PPT only completes one category
• What is the purpose of a critical values table 


  To tell us if the calculated value is significant (use 0.05)