Page 143 - Data Science class 11
P. 143
• Helps study rare events: Observational studies are helpful in studying infrequent events where we can retrospectively
gather data to determine probable causes. Also, observational studies assist researchers in knowing what to expect in
real-life situations. These studies serve as a cluster of data from standard practice.
3.2.3 Disadvantages of Observational study
The disadvantages of observational studies are as follows:
• Time commitment: The amount of time spent on an observational study may not always be justified by the findings.
• An event's non-occurrence: Certain events are uncertain and may not occur in the presence of an observer.
• Observer negligence: The observer may occasionally fail to report important observational details.
• Biased opinion: The chances of an unfair conclusion increase significantly in cases where an expert has not performed
the study’s analysis.
• Mediocre as an evidence: Observational studies are considered inferior to experimental studies, are more prone to
prejudice, and cannot be used to exhibit causality. Observational studies can be either retrospective (using existing
data) or prospective (collecting new data).
• Biased: Some of the weaknesses are human bias because the observer is a human with his/her own prejudice or bias.
Moreover, one cannot know what the subject is thinking, the subject's perspective or decision-making process.
• Bogus answers of respondents: Following are the factors on which survey data may depend: Respondents may
not feel motivated to provide precise and honest answers. Respondents may not be fully aware of their reasons or
justifications for any given answer because of lack of memory on the subject, or even boredom.
• Cost: Cost is one of the limitations of the observation method. Under most scenarios, observational data is more
expensive to obtain than other survey data. The observer remains idle while observing the events. This idle time
further adds to the cost.
• More than one control group: In general discussions of observational studies, the possibility of using more than
one control group has often been briefly mentioned. Many observational studies have even used two control groups.
However, a second control group can be of little value—even though it happens in worst of circumstances.
• Cannot show cause and effect relationship: Observational studies cannot demonstrate that the associations
identified represent cause-and-effect relationships. However, observational studies can only demonstrate that notable
associations exist between predictors and outcome variables.
Recap
Given all the information available, including the present and the historical data, forecasting can be defined as a statistical
Ÿ
task that predicts the future as accurately as possible.
An analytical study is one in which action will be taken on a cause system to enhance the performance of the system of
Ÿ
interest in future. It focuses on prediction or forecasting.
In a controlled experiment, we assign people or things to groups and apply some treatment to one of the groups, while
Ÿ
the other group does not receive the treatment.
Ÿ An observational study can be defined as a procedure in which the subjects are just observed, and the results are then
noted. During the investigation, nobody tries to interfere with the subject to affect the outcome.
In trial assessments, the subject is assigned to a random treatment and control group. However, it is unethical to expose
Ÿ
the subject to arbitrary treatment in specific scenarios.
Observational data is collected through observation of a behaviour or activity. It is collected through methods like human
Ÿ
observation, open-ended surveys, or the use of an instrument or sensor to monitor and record information, as in the case
of quantitative observation.
There are specific steps in the process of modeling that must be followed if learning is to be successful. These steps include
Ÿ
attention, retention, reproduction, and motivation.
An observation uses your five senses, whereas an inference is a conclusion based on many observations.
Ÿ
There are three different types of methodologies for conducting observational research: controlled observations, naturalistic
Ÿ
observations, and participant observations.
Forecasting on Data 141

