Statistics for ESS
T tests
This is a type of inferential statistic that is useful for comparing the means of 2 sets of data. A T test purpose is to check if the averages of 2 means is reliably different. Inferential statistics can allow us to generalise our finding to a population beyond the sample size. calculation of a T test will help you discuss your findings in a wider context.
watch these videos - it is very clear and will give you lots of things to use in your evaluation if you use this test
A large t value indicates there is a reliable difference between the means and a small t value means that the groups are similar and any difference is not reliable.
degrees of freedom - sample size -1
There are 3 types of t test
Independent samples - 2 different groups (between samples/unpaired sample) scores between groups are not related
paired samples - this is when you have one group that is measured at two different times
One sample t test - one group to compare to a known value
A good guide line is to have 20-30 data points per group - any smaller and you won't be able to find significant differences in the data.
assumptions of the students T test - you can assume;
the data is normally distributed
there is similar variance between the 2 data sets
you must have similar number in both data sets
you must have around 20-30 data points
You will perform a 2 tailed unpaired t test which can easily be performed in a few seconds in Google sheets
How to write this in your RAC section
An independent samples t test was used to check the effectiveness of ttestosterol in reducing cholesterol t(99) = 0.33, p = 0.37, but no significant diference was found (ttesterol M-34); control M = 36)
Mann Whitney U Test
You would use this if you are
You need a sample size of at least 20
Chi squared test
Chi squared test will test for patterns or relationships in categorical varibales (non numeric variable e.g gender, things you can count but aren't numbers)
Questions could be; Is there a gender balance between IB subjects: If there is a relationship we would expect an uneven spread, if not there no relationship we would expect an even spread.
One way chi squared
Two X Two design
Is the difference due to chance or is it reliable - is is another example of an inferential statistic - can you apply your results to the wider population
Spearmans Rank Correlation coefficient
This is a great test to do if you are doing the aquatic or terrestrial fieldwork and are wanting to do inferential statistics of a a possible correlation