So you have a hypothesis and now you also have some data.

What type of test should you carry out to determine whether to reject the null hypothesis?

That depends on what type of data you have and how many you have collected.

We can use a simplified scheme for determing which test we need to use

The one below is taken from Barnard et al (2007), “Asking Questions in Biology”. The book contains both this and a more detailed version.

Some examples

You have carried out a survey that asked people their preference for a range of dog grooming products.

From a range of products A-E, they had to select their favourite.

Suppose the results were:

  • Product A: 7
  • Product B: 4
  • Product C: 11
  • Product D: 8
  • Product E: 14

What kind of data are these: categorical, ordinal, constant interval or count?

  1. What might be a suitable null hypothesis in this case?

  2. Are we looking for a trend or a difference?

  3. How many groupings are there?

  4. How many treatments?

  5. Are data for each treatment replicated?

  6. Are data in the form of counts?

You have gathered pain scores for 50 cats in post-operative care that had been offered analgesic A,

and for 50 oothers that had been offered analgesic B. You want to find out if the pain is reduced for

one analgesic compared to another.

What kind of data are these: categorical, ordinal, constant interval or count?

  1. What might be a suitable null hypothesis in this case?

  2. Are we looking for a trend or a difference?

  3. How many groupings are there?

  4. How many treatments?

  5. Are data for each treatment replicated?

  6. Are the data in one treatment independent of those in other treatments?

  7. Are the data normally distributed?

You have measured body temperatures of 50 horses using both a rectal thermometer, and an infra-red

thermometer, which you pointed at the eyelid.

The infrared thermometer is non-invasive and less distressing to the horses, but you need to find out whether it reliably gives the same reading as the rectal thermometer.

For each horse, you record the readings of the two thermometers.

  1. What might be a suitable null hypothesis in this case?

  2. Are we looking for a trend or a difference?

  3. How many groupings are there?

  4. How many treatments?

  5. Are data for each treatment replicated?

  6. Are the data in one treatment independent of those in other treatments?

  7. Say again, how many treatments were there?