As with almost everything in the world of mathematical and statistical research, the field of survey coding is going to be highly dependent on the sample size in question. Consider how many people were surveyed to check the margin of error. Figuring this margin as well as a QDA miner out is going to be extremely important when it comes time to do coding.

## Margins and Survvey Coding

Results of estimation are usually expressed as a single mathematical value, which is known as a point estimate. Sometimes they’ll instead show up as a range of values. These values are collectively referred to as the confidence interval.

Whenever point estimation is used, the margin of error for the survey size is calculated. For example, consider a population proportion that was surveyed and is represented by P as a value. The margin of errors calculated could be:

*x1.97 [P(1-P)/n]1/2*

Think back to the last time you might have seen a poll conducted in a newspaper or on television news broadcast features. The margin of error often appears in a small typeface at the bottom of the actual information. Reporting only this error isn’t much information by itself. There should be some degree of confidence in the findings. The most important part is the sample size (n) that participated in the survey.

Social science studies can often be broken down into simple yes/no questions, and those questions usually have to be coded so that there’s a proportion between responses as a proportion of the P total. A simple random survey coding sample would show a P variance of:

*P(1-P)/n*

By contrast, a 95 percent confidence interval would be:

*P – 1.96 [P(1-P)/n]1/2, P + 1.96 [P(1-P)/n]1/2*

When solving for sample sizes with binary data:

*n = [t2 N p(1-p)] / [t2 p(1-p) + a2 (N-1)]*

There’s actually a pretty good rule of thumb to keep in mind in regards to questionnaire coding. Whenever finding how accurate a survey code is, consider:

*d = Absolute Precision = (reliability coefficient) .(standard error) = Z a/2 . (S/n1/2)*

This shouldn’t be taken as something that’s totally always accurate, but it should be as long as the math has all been done correctly.

## Accuracy of Surveys

Every time someone codes a survey they start to wonder, midway through after the statistical data analysis has been collected, if they have enough people and what percentage or fraction of the population is enough. This sort of debate is actually more academic than it might seem, and in practice it simply goes that the large the sample size the better the survey will be after it’s coded. It might honestly be viewed as that easy, but of course it’s not always easy to attract a larger and larger sample size to be surveyed. Just make do with the best poll options that were available at the time.