WK1 Lecture 2 - Inference and Bias
Statistical Inference
Statistical inference is the process of inferring something about a population based on information obtained from a sample. All statistical inferences make assumptions, including assumptions about the sample. If sampling is not of good quality any inferences are invalid.
Sample Quality
Good quality samples are representative of the population of interest and is ideally a random selection of members of a population. However, this is difficult to achieve in practice. Conclusions occasionally come with caveats due to the limitations of data.
Representation
Ideally a sample should be representative of the population as a whole. If a sample is representative then inferences based on the sample will give a reliable guide to the population. Good representation is more likely to occur if the sampling process is free from bias which is again difficult in practice.
Sources of Bias
Common sources of bias include:
Sampling bias - When some members of the population are systemically more likely to be selected in a sample than others.
Self-selection bias - Self selection tends to over-represent people with strong opinions.
Investigator intervention - Actions, expectation, or involvement of the 'investigator' alter results. e.g. the expectation of the "average group" may lead to a biased sample.
Systemic sampling - Can be biased based by periodical or seasonal events.
Convenience - Parts of the population that are harder to observe may be excluded.
Non-response bias - Refusers and participants that drop out may systemically differ from those who do not.
Simple random sampling by definition should avoid bias, but this relies upon a list of the population. It doesn't always ensure representation despite being a high quality data collection procedure as there's always a chance of an anomalous sample.