Professional Documents
Culture Documents
When the researched sample data results in a statistical value that does not conform to its
population parameters accurately and precisely it means a statistic error (statistic error). There
are two factors that are the cause of statistic error, namely:
The unit sampling error is a type of error caused by the determination of the elements in a
sample unit that may be less representative of its population characteristics. The level of
heterogeneity of population elements can lead to errors in sample units defined by specific
strata or clusters. Certain elements may have the opportunity to enter into multiple strata or
groups of sample units. Because if an element is only chosen once, it will likely cause an
error in classifying element into several strata or groups as the sample unit is a source of
error caused by the determination of sample units. Mistakes that occur are usually caused by
the researcher's fault in determining the sampling method.
A systematic error is a fault caused by factors outside the sampling process (non-sampling
error). This systematic error usually occurs because of the design weakness of research and
misconduct of research. There are two factors that affect the systematic error namely:
1. Respondent Error
The respondent's fault is the type of error caused by the respondent's response to the survey
method. If the research respondent wants to cooperate and answer the question correctly, then the
results of the data obtained will fulfill the expected objectives. Error respondents consist of two
types, namely:
A Non-response bias error is a mistake that occurs because the subject of a sample that does
not provide a response is more representative than the subject of the sample that provides the
response, so that the investigated samples are less accurate and precise in the reflects its
population characteristics. This issue does not only occur in the collection of data with surveys
by post or shared-link, but also happens to surveys through individuals, interviews over the
phone and face-to-face interviews. For example, in a survey, a result of 30% of respondents gave
a response and as much as 70% of respondents did not respond. Up to 70% of samples are more
representative than sample subjects that provide feedback.
Example: In a study titled “Quality Influence of Accounting information system against COSO
based on Internal control and its impact on fraud prevention”. Researchers use a Questionnaire
deployment as a data collection method. If the respondent who requested to fill the Questionnaire
replied with a choice that is neutral. For example, there are five options in the questionnaire,
namely SL (always), SR (often) K (sometimes), J (rarely), TP (never) and on all questions are
answered with the answer K (sometimes) is a mistake caused because the respondent does not
give The correct answer to the questionnaire.
A response bias error is an error that occurs because the respondent gives an incorrect answer
in reality either intentionally or unintentionally. Incorrect or inappropriate answers given by the
respondent led to erroneous researcher's interpretation of the respondent's responses. Some
things that can create a bias response, namely:
The trend of respondents who gave an answer agrees to the questions that he or she did
not understand (acquiescence bias).
Propensity of respondents that provide extreme answers (externity bias) or neutral
(neutrality bias) to most questions.
Mutual role between interviewers and respondents so that the respondent's answers are
affected by the interviewer's opinion (refractive) or the interviewers ' assistance to the
respondent to answer the question (auspice bias).
Example: In a study entitled Influence of Quality accounting information system against
COSO based Internal control and its impact on fraud prevention. Researchers use a
Kuisoner deployment as a data collection method. If the respondent who requested to fill
the Kusioner replied with a choice that is neutral. For example, there are five kinds of
options in the questionnaire namely SL (always), SR (often) K (sometimes), J (rarely),
TP (never), then the respondent replied with random without understanding the intent of
the statement in the questionnaire.
Example: In a study titled Implementation of Accounting Information system in cash
receipts. The mistake that researchers might do is that the possibility of double entering
data when using SPSS.
2. Administrative errors
Error processing data is likely to occur due to errors in procedural or arithmetic processes
through a computer. Despite using a computer, accuracy remains dependent on human precision
in the creation of programs and inserting data into the computer. However, this type of error can
be minimized by setting the procedure carefully and carefully starting from editing data, coding,
and other stages of data processing using the computer.
b. Interviewer error
Interviewer errors are administrative mistakes caused by interviewer accuracy. The error may
be an interviewer's fault in noting the respondent's response or missing part of the information
that is important because the interviewer is less quickly noting the respondent's response orally.
The interviewer's fault can also be caused by a selective perception of the interviewer who only
records the respondent's answer that is not in line with the interviewer's attitude and opinion.
Example:
Example:
The population is a sepsific population that is relevant to the purpose or problem of research.
According to Kuncoro (2010:123) The population can be distinguished into a limited population
and an unlimited population. The limited population is a very quantitative source of data, which
can be relatively numbered. The infinite population is a data source that cannot be determined so
that the relative cannot be expressed in the form of quantities.
The sample selection Framework is a list of the population elements that are used as the
basis for sampling.
Degree uniformity, if the popolation is perfectly uniform, then one elementary course
of the entire population is enough representative to be researched. If the population is
completely heterogeneous, then only the complete counting can give a representative
picture.
The desired precision in the study, the determination level is determined to be the
difference in the results of the sample differences with a complete counting result,
assuming the instrument, the interview technique, the interview quality used the same.
Quantitatively the precision measured from the Erroe standard, the smaller the default
error, the greater the level of precision.
Analysis plan, data analysis plan with specific analysis techniques greatly determines
the amount of sample to be taken
Subject to resource availability
4) Determination of sample size
The Data to be analyzed is obtained from research samples. Thus, the greater the number
or size of samples with the correct sample method of course the data obtained will be more
representative or represent the population studied. However, the sample size is closely related
to research costs. If the sample size is large, then the cost of research will also be greater.
the final step in the sample selection process is to select the necessary samples. A
sample Unit is an element or group of elements that become the basis to be selected as a
sample. Sample selection based on the sample framework can be done through one stage or
multiple procedures.
Each member of the population has the same opportunity to be selected as a sample in the
study.
Each element is selected independently of each other element and the sample is taken
with a random procedure from a sampling framework
For a limited and relatively small population size, the sampling process is relatively easy
and the larger the population size, the sampling process is increasingly complicated.
Sample draws from sample frameworks can be done with random tables, sweepstakes,
computers, and others.
Advantages and Disadvantages of simple random sampling (Kuncoro, 2010:127)
sampling techniques, including:
Advantages:
o Sample selection procedure is very easy and simplest
o Unti selection of samples of only one type
o Fault classification can be avoided
o Simply with a large graphical representation of the population
Disadvantages:
o A general overview of the population may have been unknown
o By using the same number of samples, the level of accuracy and accuracy of research
becomes lower than the selection of random stratification.
4. Stratified Sampling
In the selection of stratified sampling samples, the first step before selecting a sample is
to do a population grouping with certain criteria into multiple strata. Any element contained
in the population can only be incorporated into one of the strata. Each strata will serve as a
sample selection unit and each of the strata can be drafted in the sample selection framework.
Each element of the existing sample Unti will be selected randomly to become a sample.
According to Sugiarto (2017:150) characteristic stratified sampling, namely:
Simple is selected by selecting a random starting point and then taking each successive
element from the sample frame to be sampled.
Elements are selected from the population at the same interval by time, order, or room.