In this case the data are collected by observing some process which may be
well understood. For example Rural Tree Decline study was based on
observations made by Dr. Richard Doyle.
Other examples can be the study of hospital records to see if they
indicate that the incidence of a disease is related to blood groups, or
accident records might be used to access the effects of changing traffic
laws.
A sample survey, which is used to gain peoples’ attitudes to political
questions or consumer goods is also a type of observational study.
Problems that need to be overcome in observational studies
There are a number of problems associated with observational studies:
The main conclusion may be invalid because of the confounding
effects of uncontrolled variables. That is to say there may be no way of
knowing whether an effect observed in the data is due to change in the
variable of interest or is instead due to changes that happen to also
occur in other variables at the same time.
Estimates may be biased because of non-random sampling Truly random
sampling is often difficult to carry out, and there is a temptation to
assume that a sample that is obtained in some convenient way is
equivalent to a random sample. Unfortunately, however, it is then a very
systematic bias in the sampling procedure to distort estimates of key
parameters to such an extent that a study becomes quite worthless.
Where observations have to be taken on groups of individuals, rather
than on individuals themselves. For example dietary risk factors might
be assessed by comparing a number of communities in terms of their rates
of heart disease and the average amounts of different types of food
consumed. If high rates of heart disease occur in those communities that
have a high intake of certain types food, then it might be assumed that
eating these foods increases the risk of disease. However this is not
necessarily a valid conclusion and might be an example of what has been
called an
Example – Fluoridation and cancer
The study by Yiamouyyiannis and Burk(1977) claims to show there is
evidence that adding fluoride to drinking water is responsible for large
numbers of people dying from cancer each year in the United States. This
study was not well received by the medical and dental professions; with
most physicians and dentists agreeing that there is no evidence that
fluoridation at low doses is carcinogenic. It is also argued that the
adding of fluoride to water has a greatly beneficial effect on the dental
health of the population and should be continued. The data was based on
comparing cities where fluoridation took place compared to cities where
there was no fluoridation duirng1950’s - 1970. Closer examination of the
data showed changes in the demographics in terms of age, gender and
cultural differences between these two types of groups of cities had taken
place. Another factor that was overlooked was the location of the cities
and their proximity to pollution levels.
The sample survey
The potential problems with observational studies are greatest when the
data that is available and used in a study was collected for some other
purpose at some time in the past. However, the problems can be minimised
if a sample survey is carried out with the sole aim of collecting the
required data, using a well-designed sampling scheme.
There are times when a different type of sample survey is required because
it is not possible to know in advance where or when the individuals in a
population will be encountered. For example the size of an insect
population can be estimated by marking, releasing and recapturing insects,
and the number of polar bears in a region of the arctic can be estimated
by flying in a straight lines over the area and recording the positions of
the bears that are seen.
