Types and Sources of Errors in Statistical Data *
a. non-sampling errors and
b. sampling errors.
*
Non-sampling errors
These are errors that arise during the course of all data collection activities.
In summary, they have the following characteristics:
exist in both sample surveys and censuses data.
difficult to measure .
failure to identify the target population.
non response.
Defects in the sampling frame
This result in coverage errors.
These occur when there is an omission, duplication or wrongful inclusion of units in the sampling frame.
Omissions are referred to as ‘under coverage’ while duplications and wrongful inclusions are called ‘over coverage’.
These errors are caused by defects such as inaccuracy, incompleteness, duplication, inadequacy and out of date sampling frames.
*
Failure to Identify Target Population
*
Response
*
a. Poor questionnaire design
The content and wording of the questionnaire may be misleading and the layout of the questionnaire may make it difficult to accurately record responses.
As a rule, questions in questionnaire should not be loaded, double-barrelled, misleading or ambiguous, and should be directly relevant to the objectives of the survey.
*
Poor questionnaire design – cont’d
*
b. Interviewer bias
An interviewer may influence the way a respondent answers survey questions.
*
These arise through the respondent providing inaccurate or wrong information.
They occur because of memory biases or respondents giving inaccurate or false information when they believe that they are protecting their personal interests or integrity.
They can also arise from the way the respondent interprets the questionnaire and the wording of the answer that the respondent gives.
Careful questionnaire design and effective questionnaire testing can overcome these problems to some extent.
*
d. Problems with the survey process
*
Non-response results when data is not collected from respondents.
The proportion of these non-respondents in the sample is called the non-response rate.
Non-response can be either total or partial.
Total non-response or unit non-response can arise if a respondent cannot be contacted (because the sampling frame is incomplete or out-of-dated) or the respondent is not at home or is unable to respond because of language difficulties or illness or out rightly refuses to answer any questions or the dwelling unit is vacant.
*
Non-response - cont’d
When conducting surveys it is important to document information on why a respondent has not responded.
Partial non-response or item non-response can occur when a respondent replies to some but not all questions of the survey.
This can arise due to memory problems, inadequate information or an inability to answer a particular question/section of the questionnaire.
A respondent may refuse to answer if;
a. they find questions particularly sensitive, or if
b. they have been asked too many questions.
*
To reduce non-response, the following approaches can be used:
care should be taken in questionnaire design through the use of simple questions.
pilot testing of the questionnaire.
explaining survey purposes and uses.
assuring confidentiality of responses.
*
Processing
These occur at various stages of data processing such as data cleaning, data capture and editing.
Data cleaning involves taking preliminary checks before entering the data onto the processing system.
*
Processing – cont’d
Inadequate checking and quality management at this stage can introduce data loss (where data is not entered into the system) and data duplication (where the same data is entered into the system more than once) thus introducing errors in data.
*
Time Period Bias
*
Analysis and Estimation
Analysis errors include any errors that occur when using wrong analytical tools or when preliminary results are used instead of the final ones.
Errors that occur during the publication of the data results are also considered as analysis errors.
Estimation errors occur when inappropriate or inaccurate weights are used in the estimation procedure thus introducing errors to the data.
*
Can be minimised by adopting any of the following approaches:
using an up-to-date and accurate sampling frame.
careful selection of the time the survey is conducted.
planning for follow up of non-respondents.
careful questionnaire design.
*
Reducing non-sampling errors – cont’d
*
Sampling error
Refer to the difference between the estimate derived from a sample survey and the 'true' value that would result if a census of the whole population were taken under the same conditions.
These are errors that arise because data has been collected from a part, rather than the whole of the population.
*
Sampling errors – cont’d
There are no sampling errors in a census because the calculations are based on the entire population.
*
a. sample size.
In general, larger sample sizes decrease the sampling error, however this decrease is not directly proportional.
As a rough rule of the thumb, you need to increase the sample size fourfold to halve the sampling error but bear in mind that non sampling errors are likely to increase with large samples.
b. the sampling fraction.
*
Factors Affecting Sampling Error – cont’d
c. the variability within the population.
More variable populations give rise to larger errors as the samples or the estimates calculated from different samples are more likely to have greater variation.
The effect of variability within the population can be reduced by the use of stratification that allows explaining some of the variability in the population.
d. sample design.
*
Characteristics of the sampling error
generally decreases in magnitude as the sample size increases (but not proportionally).
depends on the variability of the characteristic of interest in the population.
can be accounted for and reduced by an appropriate sample plan.
can be measured and controlled in probability sample surveys.
*
Reducing sampling error
*