statistical testing

• what is the sign test used for?

  Used to analyse the difference in results, it can be used to determine if something is significant or not.
• when do we use the sign test?

  If a matched pairs or repeated measures design has been used and if nominal level data has been collected. Has to be an experiment. To know if we need to use the sign test look for:

  -difference

  -design

  -data
• calculating the sign test

  - put the data in a table.

  - use a + or - to show the number of scores that increased or decreased in the conditions (if there are participants who showed no difference, do not count them in the sign test).

  - the lowest value (so either + or -) becomes the value 'S'.

  - the 'S' value needs to be the same as or less than the critical value to be significant.

  - you need to calculate how many participants were in the research (n/o participants - n/o participants who did not change).

  - once you have done this you can look at the given critical value table, find the number of participants and level of probability used.

  -then look at the table and see if your 'S' value is less than or equal to the critical value. If it is the findings are significant; if the 'S' value is higher than the critical value the findings are insignificant.
• 3 decisions to make when picking a statistical test

  -difference or correlation

  -design (skip if it is a correlation)

  -type of data (if it is a correlation & there is more than one type of data pick the lower level e.g., ordinal over interval)
• design

  -related: repeated measures, matched pairs

  -unrelated: independent groups
• statistical tests & when to use them

  Testing for a difference:

  Unrelated design:

  Nominal: Chi-squared

  ordinal: Mann Whitney U test

  Interval: Unrelated T test

  Related design:

  Nominal: sign test

  Ordinal: Wilcoxon Signed Ranks test

  Interval: Related T test

  
  

  Testing for a correlation:

  Nominal: Chi squared

  Ordinal: Spearman's Rho

  Interval: Pearson's r
• parametric test (what it is & the tests)

  -powerful & sensitive test, more likely to detect a significance compared to other tests.

  -only done on interval data that is normally distributed.

  -data must show homogeneity (similar variation between the 2 groups).

  -assess the tests by comparing the standard deviations (they should be similar).

  -these tests: unrelated T test; related T test & Pearson's r
• mnemonic for remembering statistical tests

  Carrots Should Come Mashed With Swede Under Roast Potatoes.