"There are three degrees of untruth:
lies, damned lies, and statistics."

Whatever its origin, this sentiment has a grain of truth: dubious statistics are frequently used to bamboozle the unwary (see examples here). On the other hand, plenty of unconscious biases can arise if we don't look at figures properly.

A more positive approach to the topic is:

"There are three degrees of near-truth:
guesses, guestimates, and statistics."

A quick experiment

Consider the following questions:

1. How many Iban are there in the Kuching Division of Sarawak?
2. How many registered voters are there in the Bukit Selambau constituency?
3. What is the expenditure per employee of the bank HSBC Malaysia Berhad?

First, try a guess at the correct answer.

Second, get together some easily accessible information and make a better guess - a guestimate - of the number. For example:

1. The total population of Sarawak is about 2 million; Kuching is the most populous of 9 divisions; the Iban are one of the largest ethnic groups in Sarawak.
    
2. The total population of Malaysia is about 27 million and there are 219 members of parliament.
    
3. Make a rough breakdown of employee costs such as salary, pension and Social Security costs, training, etc. and add them up.

When you are satisfied that you have the best possible guestimates, click here to see the true values.

Third, consider how you would design a study to provide a better estimate of each quantity if the true answer was not available. In most cases you will need to collect data about a sample of the people involved and then make an inference about the true value. This requires statistics!

Main points:

• The true values include brief explanations of what is being measured - in particular the date - and the source of information.
   
• The true values are referred to as parameters; they are not statistics. In ecology we rarely know the true values, and must use estimates inferred from samples.
   
• Because our inferences are not exact, they must be accompanied by an indication of precision, such as Standard Errors or Confidence Intervals.

A statistic is not a 'true' value, but is an inference made in a mathematically sound manner from a sample, and accompanied by a statement of how precise we think it is.

What next?

For more on statistical inference look at the "squirrels" exercise.