Working through the Data

6.7 Knowing about inferential statistics

As already mentioned, a second level of statistical analysis can help determine how meaningful the data are, (i.e., how well they represent reality, rather than being the result of chance) and how well they might pertain to another population.

Such statistical analysis can be done by community groups but how and when it should be done depends on a variety of factors related not just to the data, but to the whole research design, e.g., the type of data, number and selection of respondents and their characteristics compared to the larger population. Such analysis should be done by staff, volunteers or community partners with specific knowledge and experience who can help design data collection from the outset.

However, it is helpful to have some knowledge of the jargon to understand what experts are able to provide and to aid in assessing research related to similar programs or interests.

The most commonly accepted way to express the odds of some finding in the data being true (an accurate reflection of reality), rather than the result of chance, (i.e., the finding was the result of some other incidental factors) is the p-value. Individual evaluators can decide what level of p-value is appropriate for the data being investigated but generally p<0.05 is acceptable. This means there is only 1 chance in 20 that the result you see from the data is accidental. Any results meeting the p<0.05 level of probability are called ‘statistically significant’ and those with smaller p-values (e.g. p<0.001) are ‘statistically highly significant.’ Data without that p-value may still be accurate but need to be viewed with more caution.