5. Problems cause by coding data

Whilst coding data can bring many benefits it can also lead to some problems.

Coarsening of data

This means that during the coding process some of the subtle details in the data are lost.

Look at the image below:

The colours of the houses could be classed as:

Light pink, pale blue, black and mid blue

However, when these colours are coded they may become:

PK (pink), B (blue), BK (black), BE (blue)

In this case, no allowance has been made for shades of colour so the results from the above coding would end up as this:

The fine detail have been lost. This is what is meant by 'coarsening of data'.

Coding can obscure the meaning of the data

A reader seeing the 'gender' data as M/ F is pretty likely to know that it means Male/ Female.

But some codes are more obscure, for example the country code for Switzerland is CHE. Many people might not recognise what this code represents.

If you were given the code, 244/5838 would you know what this represented? Have a search on the Argos site to see if you can find this product.

In order for the code to be useful, you need to be given a complete list of possibilities.

Coding of Value Judgments

When you are collecting data about people's opinions it might be difficult to code their answers with accuracy.

For example, you might you ask the question, "was that curry too spicy?". Your plan is to give their answers a code from 1-4 with 1 being mild to 4 being 'blow your head off'. However, what is spicy to one person will be mild to another. The code they give will depend on their individual opinion.

Coding of value judgments will inevitably lead to coarsening of the data since there will be a wide range of opinions that could be held and only a limited number of codes available.