Vous êtes sur la page 1sur 13

Probability (random) sampling methods

Probability sampling methods have been developed to ensure some


degree of precision in estimations of the population parameters. Thus,
probability samples reduce sampling error. The term probability sampling
methods refers to the fact that every member (element) of the population
has a probability higher than zero of being selected for the sample.
Inferential statistical analyses are based on the assumption that the
sample from which data were derived has been obtained randomly. Thus,
probability sampling method are often referred to as probability sampling
methods. These samples are more likely to represent the population than
are samples obtained with nonprobability sampling methods. ll subsets
of the population, which may di!er from one another but contribute to the
parameters of the population, have a chance to be represented in the
sample. Probability sampling methods are most commonly used in
"uantitative and outcomes research.
There of less opportunity for systematic bias if sub#ects all selected
randomly, although it is possible for a systematic bias to occur by chance.
$sing random sampling, the researcher cannot decide that person % will
be a batter sub#ect the study than person &. In addition, a researcher
cannot e'clude subset of people from selection as sub#ect because he or
she does not agree with them. (oes not like them, or )nds them hard to
deal with. Potential sub#ect cannot be e'cluded because they are too sick,
not sick enough, coping too well, or not coping ade"uately. The
researcher, who has a vested interest in the study, could ( consciously or
unconsciously ) select sub#ect whose conditions or behaviors are
consistent with the study hypothesis. It is tempting to e'clude
uncooperative or assertive individuals. *andom sampling leaves the
selection to chance and, thus, increases the validity of the study.
Theoretically, to obtain a probability sample, the research must
develop a sampling frame that includes every element in the population.
The sample must be randomly selected from the sampling frame.
Simple random sampling
+imple random sampling is the most basic of the probability
sampling methods. To achieve sample random sampling, element are
selected at random from the sampling frame. This goal can be
accomplished in a variety of ways, limited only by the imagination of the
researcher. If the sampling frame is small, the researcher can write names
on slips of paper, please the name in a container, mi' well, and then draw
out one at a time until desired sample size has been reached. nother
techni"ue is ti assign a number to each name in the sampling frame. In
large population sets, element may already have assigned to medical
records, organizational memberships, and licenses.
The can be some di!erences in the probability for the selection of
each element, depending on whether the selected element,s name or
number is replaced before the ne't name or number is selected. +election
with replacement, the most conservative random sampling approach,
provides e'actly e"ual opportunities for each element to be selected
(-erlinger . le, /000).
1lynn (/002,p./00) used sample random sampling in her study that
3e'plored association between organizational support for nursing practice
in home health care agencies and (a) the fre"uency of nurse4reported
adverse events, (b) nurse assessed "uality of care, (c) nurs #ob
satisfaction, and (d) nurses intension to leave their employing agency.
Stratifed random sampling
+trati)ed random sampling is used when the researcher knows
some of the variable in the population that are critical to achieving
representativeness. 5ariable commonly used for strati)cation are age,
gender, ethnicity, socioeconomic status, diagnosis, geographical region,
type of institution, type of care, care provider, and site of care.
$lrich et al. (/006) used a strati)ed random sampling method to
obtain their sample of nurse practitioners (7Ps) and physician assistant
(Ps) for the purpose of studying the ethnical con8ict of these health care
providers associated with managed care. The following e'cerpt from this
study describes the sampling method used to obtain the )nal sample of
9:;6 providers (<;; 7Ps and 6<= Ps.)
Cluster sampling
>luster sampling is a probability sampling method applied when the
population is heterogeneous? it is similar to strati)ed random sampling but
takes advantage of the natural clusters or groups of population unit that
have similar characteristic (fawcett . @arity, /00=). >luster sampling is
used in two situation is when a simple would be prohibitive in the term of
travel time and cost. Imaging trying to arrange personal meeting with900
people, each in a di!erent part of the united states.the second situation is
in case in wich the individual elements making up the population are
unknown, preventing the development of a sampling frame. 1or e'ample,
there is not list of all the heart surgery patient who complete rehabilitation
programs in the united states. In these cases, it is often possible to obtain
list of institutions or organization with which the element of interest are
associated.
In cluster sampling, the researcher develops a sampling frame that
includes a list of all the states, cities, institution, or organizations with
wich element of the identi)ed population would be linked. states, cities,
institution, or organizations are selected randomly as units from which to
obtain elements for the sample. In some cases, this random selection
continues through several stages and is referred to as multistage cluster
sampling.
Systematic sampling
+ystematic sampling can be conducted when are ordered list of all
members of the population is available. The process involves selecting
every individual on the list, using a starting point selected randomly. If the
initial starting point is not random, the sample is not a probability sample.
Tu use this design in your research, you must know the number of element
in the population and the size of the sample desired. (ivide the population
size by the desired sample size, giving k, the size of the gap between
elements selected from the list.
k : population size / sample size
NONPROBABII!" (NONRAN#O$) SA$PIN%
$&!'O#S IN ()AN!I!A!I*& R&S&ARC'
In non probability sampling, not every element of the population has
an opportunity to be included in the sample. Thus, non probability
sampling methods increase the likelihood of obtaining samples that are
not representative of their target population. Aow ever, the ma#ority of
nursing studies use non probability probability sampling, especially
convenience sampling, to select study samples. n analysis of nursing
studies published in si' nursing #ournals from 9=22 to 9=<6 revealed that
=9 used random sampling, which is a trend that appears to continue to
day (moody et al.,9=<<)
There are several types of non probability (nonrandom) sampling
designs. Bach addresses a di!erent research need. The )ve non
probability sampling design described in this te'tbook are (9) convenience
sampling, (/) "uota sampling, (;) purposive sampling, (C) network
sampling, and (:) theoretical sampling.
Con+enience (accidental) sampling
In convenience sampling, sub#ects are included in the study because
they happened to be in the right place at the right time. *esearch simply
enter available sub#ect in to the study until they have reached the desired
simple size. >onvenience sampling, also called accidental sampling, is
considered a weak approach to sampling because it provides little
opportunity to control for biases. Dultiple biases may e'ist in convenience
sampling ? these biases range from minimal to serious.
*esearch must identify and describe know biases in their simples.
&ou can identify biases by carefully tinking trough the sample criteria used
to determine the target population and the taking stapes to improve the
representativeness of the sample.
Dany strategies are available for selecting a convenience sampling.
classroom of student might be used. Patients who attend a clinic on
speci)c day, sub#ects who attend a support group, patient currently
admitted to a hospital with a speci)c diagnosis, and every )fth person
who enter the emergency department are e'amples of types of commonly
selected convenience samples.
>onvenience sample are ine'pensive and accessible, in they usually
re"uire less time to ac"uire than other types of samples. >onvenience
sample provide means to conduct studies on topics that could not be
e'amined through the use of probability sampling. Thus, >onvenience
sample enables researchers to ac"uire information in une'plored areas.
ccording -erlinger and Eee (/000), a convenience sample is probably not
that bad when it is used with reasonable knowledge and care in
implementing a study.
(uota Sampling
Fuota sampling used a convenience sampling techni"ue with an
added feature, a strategy to ensure the inclusion of sub#ect types that are
likely to be underrepresented in the convenience sample, such us women,
minority groups, the elderly, the poor, the rich, and the undereducated.
This method may also be used to mimic the known characteristic of the
target population or to ensure ade"uate number of sub#ect in each
stratum for the planned statistical analyses. The techni"ue is similar to
that used in strati)ed random sampling. If necessary, mathematical
weighting can be used to ad#ust sample values so that they are consistent
with the proportion of subgroups found in the population.
Dc>ain et al. (/00;) used "uota sampling to study the e!ect of
stress management on the psychoneuroimmunology outcomes in person
with AI5 infection. They described their sample selection as follow.
Nonprobability sampling methods use in ,ualitati+e
research-
Fualitative research is conducted to gain insight and discover
meaning about particular e'perience, situation, culture element, or
historical event. The intent is an in4depth understanding of a purposefully
selected sample and not the generalization of the )ndings from a
randomly selected sample to a target population, as in "uantitative
research.
Three common sampling methods used in "ualitative research are
purposive sampling, snowball sampling, and theoretical sampling.
(Darshall and *ossman, /006)
Purposive sampling
In purposive sampling sometime referred to as judgmental or
selective sampling, the researcher consciously select certain
participants, elements, events, or incident to include in the study. In
purposive sampling, "ualitative researchers select information4rich
cases, or those cases that can teach them a great deal about the
central focus or purpose of the study. In purposive sampling,
"ualitative researchers select information4rich cases, or those cases
that can teach them a great deal about the central focus or purpose of
the study (@reen . Thorogood, /00C? Patton, /00/).
This sampling plan has been criticized because it is diGcult to
evaluate the precision of the research,s #udgment. Aow does one
determine that the patient or element was typical or atypical, good or
bad, e!ective or ine!ectiveH Thus, research need to indicate the
characteristics that they desire in participant and provide a rationale
for selecting these types of participant to obtain essential data for
there study.
7etwork sampling
7etwork sampling sometimes to as 3snow ballingI, hold promise
for location sample diGcult or impossible to obtain in other ways or
who had not been previously identi)ed for study. 7etwork sampling
takes advantage of social networks and the fact that friend tend to
have characteristic in common. Jhen you have found a few
participants with the necessary criteria, you can ask for their
assistance in getting in touch with other with similar characteristic.
The )rst few participants are often obtained through a convenience
sampling method, and the sample size is e'pended using network
sampling (Darshall . *ossman, /006? Dunhall, /009).
In "ualitative research, network sampling is an e!ective strategy
for identifying participants who know other potential participants who
can provide the greatest insight and essential information about an
e'perience or event that this identi)ed for study (Patton, /00/)
Theoretical sampling
Theoretical is usually used in grounded theory research to
advance the development of a selected theory throughout the
research process (Dunhall, /009). The research gathers data from any
individual or group that can provide relevant data for theory
generation. The data are considered relevant if they include
information that generates, delimits, and saturates the theoretical
codes in the study needed for theory generation. code is saturated if
it is complete and the research can see how it )ts in the theory. Thus,
the research continue to seek sources and gate data until the codes
are saturated and the theory evolves from the codes and the data.
(iversity in the sample is encouraged so the theory developed covers
a wide range of behavior in varied situation (Paton,/00/).
*ew (/00;) conducted a grounded theory study develop a theory
of taking care of oneself that was grounded in the e'periences of
homeless youth. The study incorporated theoretical sampling, and the
sampling method was described as follows.
Sample si.e in ,uantitati+e research
n increasing number of nurse research are using power analysis to
determine simple size, but it is essential that the results of the power
analyses be included in the published studies. 7ot conduction a power
analysis for study or omitting the power analysis result in a published
study are signi)cant di!erences or relationship, which might be due to an
inade"uate sample size.
Earge sample size are diGcult to obtain in nursing studies, re"uire
long data collection period, and are costly. Therefore, in developing the
methodology for a study, you must evaluate the element of the
methodology that a!ect the re"uired sample size. -raemer and thiemann
(9=<2) identi)ed the following factors that must be taken into
consideration in determining sample size K
9. The more stringent the signi)cance level (e.g., 0.009 versus 0.0:), the
greater the necessary sample size.
/. Two4tailed statistical test re"uire large sample size then one tailed
test.
;. The smaller the e!ect size, the larger the necessary sample size
C. The larger the power re"uired, the larger the necessary sample size
:. The smaller the sample size, the smaller the power of the study.
B!ect size
B!ect is the presence of a phenomenon. If a phenomenon e'ists, it
is not absent, and thus, the null hypothesis is in error. B!ect size (B+) is
the e'tent of the presence of a phenomenon in a population. B!ect, in this
case, is used in a broader sense than that of 3 cause in e!ectI.
B'ample, if the mean weight loss for the treatment or intervention
group is : pounds per month with a standard deviation (+() of C.: and the
mean weight loss of the control or comparison group is 9 pound per month
with a +( L 6.:, you can calculate the B+
B+ L : 4 9 K 6.: L C K 6.: L 0.69: L 0.6/
B+ L Dean of the treatment group 4 mean of the control
group K standard deviation
of control or comparison group
Type of study
(escriptive case studies tend to use small sample. @roup are not
copered, and problem related to sampling error and generalization have
little relevance for such study. small simple size may batter serve the
research who is interested in e'amining a situation in depth from various
perspectives. Mther descriptive studies, particularly those using survey
"uestionnaires, and correlational studies often re"uire large sample. In
these studies, multiple variable may be e'amined, and e'traneous
variables are likely to a!ect sub#ect responses to the variables under
study. +tatistical comparison are often made among multiple subgroups in
the sample, re"uiring that an ade"uate ample be available for each
subgroup being analyzed. In addition, sub#ect are likely to be
heterogeneous in term of demographic variables, and measurement tools
are sometime not ade"uately re)ned. lthough target population may
have been identi)ed, sampling frames may not be available, and
parameters have not usually been well de)ned by previous studies. ll of
these factors lower te power of the study and re"uire increases in sample
size (-raemer . Thiemann, 9=<2).
7umber of variables
s the number of variables under study grows, the needed sample
size may also increase.adding variables such as age, gender, ethnicity,
and education to the analysis plan (#ust to be on the safe side) can
increase the sample size by a factor of : to 90 is the selected variables
are uncorrelated with the dependent variable. In this case, instead of a
sample of :0, you may need a sample of /:0 to :00 if you plan to use the
variables in the statistical analyses.
Deasurement sensitivity
Jell4developed instruments measure phenomena with precision.
thermometer, for e'ample, measures body temperature precisely. Tools
measuring psychosocial variables tend to be less precise. Aowever, a tool
with strong reliability and validity tends to measure more precisely than a
tool that is less well developed. 5ariance tends to be higher in a less well4
developed tool than in one that is well4developed. n instrument with a
smaller variance is preferred because the power of a test always
decreases when within4group variance increases (-raemer . Thiemann,
9=<2).
(ata analysis techni"ues
(ata analysis techni"ues very in their ability to detect di!erences in
the data. +tatisticians refer to this as the power of the statistical analysis.
1or your data analysis, choose the most powerful statistical tes
appropriate to the data. Mverall, parametric statistical analyses are more
powerful than non parametric techni"ues in detecting di!erences and
should be used if the data meet criteria for parametric analysis. In many
cases, however, nonparametric techni"ues are more powerful if your data
do not meet the assumptions of parametric techni"ues. Parametric
techni"ues very widely in their capacity to distinguish )ne di!erences and
relationship in the data.
Sample si.e in ,ualitati+e research
Fualitative research, the focus is on the "uality of information
obtained from the person, situation, event, or documents sample versus
the size of the sample (Patton, /000? +andelowsky, 9==:). The sample
size and sample plan are determined by the purpose of the study. Thus,
the sample size re"uired is determined by the depth of information that is
needed to gain insight into a phenomenon, describe cultural element,
develope a theory, or describe a historical event. the sample size can be
too small when the data collected lacks ade"uate depth or richness. Thus,
an ade"uate sample size can reduce the "uality and credibility of the
research )nding (+andelowsky, /000)
The number of participants in a "ualitative study is ade"uate in
saturation of information is achieved in the study area. Important factors
that must be considered in determining sample size to achieve saturation
of data are (9) scope of the study, (/) nature of the topic, (;) "uality of the
data, (C) study design (Darshal . *ossman, /006? Dorse, /000? Dunhall,
/009? Patton, /00/)
+cope of the study
If the scope of your study is board, you will need e'tensive data to
address the study purpose, and it will take longer to reach data saturation.
Thus, a study with a purpose that has a board scope will re"uire more
sampling of participant, event , or document than would a study with a
narrow scope (Dose, /000)
7ature of the topic
If the topic of your study is clear and the participant can easily
discuss, fewer individuals are needed to obtain the essential data. If the
topic diGcult to de)ne and awkward for people to discus, however, you
will probably need a larger number of participant to saturate the data
(Dorse, /000? Patton, /00/).
Fuality of the data
The "uality of information obtained from a interview, observation ,
or document review in8uences the sample size. The higher the "uality and
richness of the data, the fewer participant you will need to saturate data
in the area of study. Fuality data are best obtained from articulate, well4
informed, and communicative participant (+andelowsky, 9==:).
+tudy design
+ome studies are designed to increase the number of interviews
with participant. The more interviews conducted wit a participant, the
greater the "uality of the data collected. 1or e'ample, a study design that
includes an interview both before an event and after the event would
produce more data than a single interview. (esigns that involve
interviewing a family or a group of individual produce more data than an
interview with a single study participant. In critically appraising a
"ualitative study, determine if the sample size in ade"uate for the design
of the study.
Research setting
The setting is the location where a study is conducted. There are three
common setting for conducting nursing research K natural, partial
controlled, and highly controlled.
7atural setting
natural setting, or )eld setting, is an uncontrolled, real life
situation or environment (Peat et al, /00/)
Partial controlled
partial controlled setting is an environment that the researcher
manipulates or modi)es in some way. n increasing number of nursing
studies, usually correlational, "uasi4e'perimental, and e'perimental
studies, are being conducted in partially controlled setting. *ivers et
al. (/00;), in a study that was introduced earlier in the discussion of
sampling frame, conducted a predictive correlational study to
determine predictor of nurses acceptance of an intravenous catheter
safety device.
Aighly controlled
highly controlled setting is an arti)cially constructed
environment developed for the sole purpose of conducting research.
Eaboratories, research of e'perimental centers, and test units in
hospitals or other health care agencies are highly controlled setting
where e'perimental studies are often conducted.
Recruiting and retaining sub/ect
*ecruiting sub#ect
The e!ective recruitment of sub#ect is crucial to the success of a
study. 1ew study have e'amined the e!ectiveness of various strategies of
sub#ect recruitment. Dost information available to guid researchers comes
from the personal e'perience of skilled researchers.some of the factor
that in8uence a sub#ect,s decision to participate in the study are the
attitudes and ethics of the researchers. The sub#ect,s need for a
treatment, the sub#ect,s interest in the study topic, fear of the unknown,
time and travel constraints, )nancial compensation, and the nature of the
informed consent (Dadsen et al., /00/? Papadopoulus . lees, /00/?
+ullivan4Nolyai et al.,/002)
The researchers initial approach to a potential sub#ect usually
strongly a!ects his or her decision about participating in the study.
Therefore, your approach must be pleasant, positive, informative, and
nonaggressive.
If a potential sub#ect refuses to participate in a study, you must
accept the refusal gracefully4in terms of body language as well as word.
&ou actions can in8uence the decision of other potential sub#ect who
observe or hear about the encounter. +ub#ect in which a high proportion of
individuals refuse to participate have a serious validity problem. The
sample is likely to be biased, because often only a certain type of
individual has agreed to participate. Therefore, keep records of the
numbers of person who refuse and, if possible, their reasons for refusal.
Jith this information, you can include the refusal rate in the published
research report with the reason for refusal. It would also be helpful if you
could determine if the potential sub#ect who refused participate di!er from
those who agreed to participate in the study. This information will help you
to determine the representativeness of your sample.
*etraining sub#ect
Mne of the serious problem in many study is sub#ect retention.
Mften, sub#ect attrition cannot be avoided. +ub#ect move, die, or withdraw
from a treatment. If you must collected data at several points over time,
sub#ect mortality can become a problem. +ub#ect who move fre"uently
and those without phones pose a particular problem. number of
strategies have been found to be e!ective in maintaining the sample. 1or
instance, it is a good idea to obtain the names,addresses, and phone
numbers (cell and home number is possible) of at least two family
members or friends when you enroll the sub#ect in the study. sk if the
sub#ect would agree to give you access to unlisted phone numbers in the
event that the sub#ect changes his or her number.
ProbabilitasOkemungkinanP (random) metode pengambilan sampel
Detode probability sampling telah dikembangkan untuk memastikan
beberapa tingkat presisi dalam estimasiQperkiraanR parameter populasi.
(engan demikian, sampel probabilitas mengurangi kesalahan sampling.
Detode probability sampling meru#uk pada fakta bahwa setiap anggota
(elemen) dari populasi memiliki probabilitas lebih tinggi dari nol untuk
terpilih sebagai sampel. nalisis statistik inferensial Qdapat disimpulkanR
didasarkan pada asumsi bahwa sampel dari data yang berasal telah
diperoleh secara acak. (engan demikian, metode probability sampling
sering disebut sebagai metode probability sampling. +ampel ini lebih
mungkin untuk mewakili populasi daripada sampel yang diperoleh dengan
metode pengambilan sampel non probabilitas. +emua subset dari
populasi, yang mungkin berbeda satu sama lain tetapi kontribusi terhadap
parameter populasi, memiliki kesempatan untuk terwakili dalam sampel.
Detode pengambilan sampel probabilitas yang paling sering digunakan
dalam penelitian kuantitatif dan hasil.
da kurang kesempatan untuk bias sistematis #ika subyek semua dipilih
secara acak, meskipun ada kemungkinan untuk bias sistematis ter#adi
secara kebetulan