-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 1
Dening and Measuring Variables
Amiel Nazer C. Bermudez, MD, MPH
Learning ObjecBves
At the end of the lecture, parBcipants should be able to:
DierenBate the following: Constant versus variable QuanBtaBve
versus qualitaBve variable Discrete versus conBnuous Scales of
measurement Independent versus dependent versus extraneous
variables
DierenBate concept, indicator, and variable Contrast conceptual
and operaBonal deniBons Enumerate suggested components of an
operaBonal deniBon
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 2
Learning ObjecBves
At the end of the lecture, parBcipants should be able to:
(con%nued)
Enumerate commonly-used methods of collecBng data Enumerate
basic principles in the construcBon of quesBonnaire Cite
commonly-used convenBons in the construcBon of quesBonnaires
Contrast open- and close-ended quesBons Discuss basic
consideraBons in the pre-tesBng of data collecBon tools
Variables
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 3
Constant versus Variable
Constant A phenomenon whose value remains the same from person
to person, from Bme to Bme, or from place to place.
Examples: speed of light, Avogadros number Variable
A characterisBc whose value diers from one individual to
another, or from one period to another in the same individual.
Examples: ages of gestaBon, smoking habit
Daniel & Cross, 2013
QuanBtaBve versus QualitaBve Variable
Qualita9ve One whose categories are simply used as labels to
dis9nguish one group from another.
Examples: sex, marital status, educaBonal a[ainment
Quan9ta9ve
One whose categories can be measured and ordered according to
quan9ty or amount, or whose values can be expressed
numerically.
Example: height, weight, number of term pregnancies
Daniel & Cross, 2013
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 4
QuanBtaBve Variable Discrete versus Con%nuous
Discrete quan9ta9ve Variable that can assume only integral
values or whole numbers
Examples: number of hospital beds, household size Con9nuous
quan9ta9ve
Variable that can assume any value including frac9ons or
decimals
Examples: serum uric acid levels, body mass index
Mendoza et al, 2010
Measurement
Dened as the assignment of numbers (or categories) to objects or
events according to a set of rules.
Since measurement may be carried out under dierent sets of
rules, there are several scales of measurement (i.e. nominal,
ordinal, interval, and raBo)
Daniel & Cross, 2013
Image Credit
h7p://cdn.theatlan%c.com/sta%c/mt/assets/business/Screen%20Shot
%202012-02-29%20at%203.08.23%20PM.png
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 5
Scales of Measurement
Nominal Categories represent a set of mutually exclusive and
exhaus9ve classes to which individuals or objects (a[ributes) may
be assigned.
Examples: sex, naBonality, blood groups, religion Ordinal
Similar with the nominal scale + categories can be ranked or
ordered
Examples: educaBonal a[ainment, severity of disease
Mendoza et al, 2010
Scales of Measurement
Interval Similar with the ordinal scale + exact distance between
all adjacent categories are equal but the zero point is arbitrary
or meaningless.
Examples: temperature measurement, calendar Bme Ra9o
Similar with the raBo scale + zero point is meaningful (i.e.
zero means absence of the a[ribute)
Examples: blood pressure, number of DMF teeth
Mendoza et al, 2010
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 6
Scales of Measurement
From the slides of Dr NR Juban, undated
Scales of Measurement
From the slides of Dr NR Juban, undated
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 7
Indicate whether the following is a qualitaBve or a quanBtaBve
variable. Indicate also the scale of measurement.
Hospital bed capacity QuanBtaBve, discrete, raBo
EducaBonal a[ainment QualitaBve, ordinal
Mid-upper arm circumference QuanBtaBve, conBnuous, raBo
Forced expiratory volume QuanBtaBve, conBnuous, raBo
Region of residence QualitaBve, nominal
ClassicaBon of Variables in Health Research
Independent variable Factor supposed to be responsible for
bringing about a change in a phenomenon or situa9on
Also called the exposure variable Dependent variable
Factor that changes with the introduc9on of an independent
variable
Also called the outcome variable
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 8
ClassicaBon of Variables in Health Research
Extraneous variables Extraneous variables may distort the true
rela9onship between the independent and dependent variables
Data on extraneous variables should be collected in the
research, to the extent possible, to allow for proper staBsBcal
adjustment.
Some types of extraneous variables Confounders Eect measure
modiers Intermediates Colliders
Extraneous Variables [opBonal] Confounder
Can result in distor9on of the true measure of eect between the
exposure and outcome.
IdenBcaBon of confounders in research is based primarily on
literature search.
ProperBes It is an independent risk factor for the outcome /
disease It is associated with the exposure It does not lie along
the causal pathway between exposure and outcome (i.e. not an
intermediate)
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 9
Extraneous Variables [opBonal] Confounder (example)
SES is a confounder because: SES is a risk factor for Kawasaki
Disease (Bronstein et al, 2000) Crowding is associated with SES
(Adler & Newman, 2002) SES is not an intermediate between
crowing and Kawasaki Disease
If SES is not properly accounted for in the study, there might
be overesBmaBon of the eect of crowding on the risk for KD
Crowding Kawasaki Disease
Socio-economic status
Variable Selec9on and Deni9on
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 10
SelecBng Study Variables
Variables to be measured in the proposed research should be
selected on the basis of their relevance to the study
objecBves.
In the face of nancial, logisBcal, ethical, or other limitaBons,
other variables that approximate the desired variable (or concept)
can be selected. For example, instead of immune status (measured by
serum anBbody levels), immunizaBon status (measured by immunizaBon
card) can be selected instead.
OperaBonal DeniBon Concept and Variable
CONCEPT VARIABLE SubjecBve impression No un i f o rm i t y a s t
o i t s understanding among dierent people
Cannot be measured
Measurable though the degree of precision varies from scale to
scale, and/or from variable to variable
Kumar R, 2011
Concept( Indicator( Variables(Concept( Indicator( Variables(
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 11
Conceptual versus OperaBonal DeniBon
CONCEPTUAL
DicBonary deniBon DeniBon widely accepted in the scienBc
community or those set forth by relevant agencies
OPERATIONAL
Meaning of the variable as specied by the researcher
Must include the acBviBes ( o r o p e r a B o n s ) f o r
measuring the variable
OperaBonal DeniBons Suggested Approach
Conceptual deniBon of the variable (may be omi[ed for common
variables)
How the variable will be measured in the study Categories of the
variable (or values the variable is expected to assume)
Method of data collecBon Possible sources of bias
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 12
OperaBonal DeniBon (example) Variable: Kawasaki Disease
Status
Opera9onal Deni9on: Diagnosed in a child who presents with fever
(T 37.7 OC) for at least 5 days AND the presence of at least four
of the ve following signs, and whose illness cannot be explained by
other known disease process: (1) bilateral bulbar conjuncBval
injecBon, generally non-purulent; (2) changes in the mucosa of the
oropharynx, including injected pharynx, injected and / or dry
ssured lips, strawberry tongue; (3) changes in the peripheral
extremiBes, such as edema and / or erythema of the hands or feet in
the acute phase; or periungual desquamaBon in the subacute phase;
(4) rash, primarily truncal; polymorphous but non-vesicular; (5)
cervical adenopathy, 1.5 cenBmeters, usually unilateral
lymphadenopathy.
(conBnued)
OperaBonal DeniBon (example) Variable: Kawasaki Disease
Status
Opera9onal Deni9on: Diagnosed in a child who presents with fever
(T 37.7 OC) for at least 5 days AND the presence of at least four
of the ve following signs, and whose illness cannot be explained by
other known disease process: (1) bilateral bulbar conjuncBval
injecBon, generally non-purulent; (2) changes in the mucosa of the
oropharynx, including injected pharynx, injected and / or dry
ssured lips, strawberry tongue; (3) changes in the peripheral
extremiBes, such as edema and / or erythema of the hands or feet in
the acute phase; or periungual desquamaBon in the subacute phase;
(4) rash, primarily truncal; polymorphous but non-vesicular; (5)
cervical adenopathy, 1.5 cenBmeters, usually unilateral
lymphadenopathy.
(conBnued)
In this example, this segment presents the conceptual deni%on of
Kawasaki Disease
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 13
OperaBonal DeniBon (example) Variable: Kawasaki Disease
Status
(conBnued)
Opera9onal Deni9on: In the study, a child is considered to have
been diagnosed with Kawasaki Disease if: (1) such diagnosis is made
as a discharge diagnosis and is indicated in the childs medical
records AND (2) the childs admiong clinical history is consistent
with the diagnosBc criteria for Kawasaki Disease.
Categories: 0 = Without the disease; 1 = With the disease Data
Collec9on: Review of discharge diagnosis AND admiong clinical
history will be reviewed from medical records.
OperaBonal DeniBon (example) Variable: Kawasaki Disease
Status
(conBnued)
Opera9onal Deni9on: In the study, a child is considered to have
been diagnosed with Kawasaki Disease if: (1) such diagnosis is made
as a discharge diagnosis and is indicated in the childs medical
records AND (2) the childs admiong clinical history is consistent
with the diagnosBc criteria for Kawasaki Disease.
Categories: 0 = Without the disease; 1 = With the disease Data
Collec9on: Review of discharge diagnosis AND admiong clinical
history will be reviewed from medical records.
while this segment presents the opera%onal deni%on.
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 14
OperaBonal DeniBon (example) Variable: Crowding
Opera9onal Deni9on: Dened as the number of persons per room in
the childs residence. The variable crowding will be determined by
means of the quesBons: How many people live in your household? How
many room divisions do you have in your house?
Categories: Enter response as is. Categories will be determined
aser data collecBon is done
Data Collec9on: Face to face interview with the childs parents
or caregiver
OperaBonal DeniBon (example) Variable: Age
Opera9onal Deni9on: The age of the child from the date of birth
to the date of hospital admission in years, months and days.
Categories: Enter response as is. Categories will be determined
aser data collecBon is done
Data Collec9on: The age will be determined through the childs
birthdate based on the records and shall be computed by nding the
dierence in year and months from the date of birth to the date of
admission.
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 15
OperaBonal DeniBon (example) Variable: Annual per capita
household income
Opera9onal Deni9on: In the study, socio-economic status will be
measured by the annual per-capita household income of the childs
household in at least one preceding year. The per-capita household
income will be determined by collecBng informaBon on the number of
household members and the combined annual income of all earning
members of the household. Earning members non-earning members of
the household will be recorded
Categories: Enter response as is. Categories will be determined
aser data collecBon is done
Data Collec9on: Face to face interview with the childs parents
or caregiver
OperaBonal DeniBon (example) Variable: WasBng
Opera9onal Deni9on: Refers to the degree of the childs wasBng,
should there be any expressed as z-scores. In the study, the
z-scores will be determined using the WHO Antro and WHO Anthro Plus
sosware.
Categories: 1 = Normal (z-score: 0.85 1.10) 2 = High WFH
(z-score: > 0.85) 3 = Low WFH (z-score: < 1.10) 99 = No
informaBon
Data Collec9on: WFH will be determined through reviewing the
hospital chart of the child and obtaining the weight and height
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 16
Comment on the operaBonal deniBon of the following:
For a research proposal that examines the relaBonship between
consumpBon of dark chocolate and pre-eclampsia, the outcome
variable was dened as:
A blood pressure of 140/90 obtained aUer the 20th weeks of
pregnancy in a pregnant woman who had normal BP before the 20th
week.
Comment on the operaBonal deniBon of the following:
For a research proposal that examines the relaBonship between
social contact factors and iniBaBon of tobacco use in adolescents,
peer cigareTe smoking was dened as:
The variable peer cigare7e smoking refers to whether the
respondent has at least one of his / her close friends smoke
cigare7es in the preceding 6 months.
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 17
Comment on the operaBonal deniBon of the following:
For a research proposal that examines risk factors for household
food insecurity, monthly household income was dened as:
This variable is obtained by dividing the combined monthly
income of all earning members of the households by the total number
of household members.
The categories of the variables are:
0 = PhP 1,403.40 1 = < PhP 1,403.40
Methods of Data Collec9on
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 18
Methods of data collecBon
Query Face-to-face interview Self-administered quesBonnaire
Guided self-administered quesBonnaire Telephone interview
Computer-assisted interview
Test administraBon IQ test Knowledge quesBonnaire
Methods of data collecBon
Group processes Focus group discussion Nominal group process
Delphi technique (consensus building)
ObservaBon Direct observaBon / ParBcipant observaBon Clinical
and laboratory examinaBons Physical / chemical / environmental
measurements Use of technology such as GIS, GPS
Review of records / documents
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 19
Steps in developing the data collecBon tool
Specify study objecBves IdenBfy study variables OperaBonally
dene study variables Specify sources of data IdenBfy method of data
collecBon for each data source IdenBfy possible sources of errors
in measurement
Steps in developing the data collecBon tool
IdenBfy data collecBon tool(s) to be used Prepare quesBons /
data forms Format / Organize quesBons Translate and back
translate
Pre-test Finalize and reproduce
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 20
Guidelines in construcBng quesBonnaires
Sequencing / OrganizaBon Formaong Phrasing
Guidelines in construcBng quesBonnaires Sequencing /
Organiza%on
Sequence quesBons that they may ow in a logical order
Suggested steps List the quesBons Group the quesBons in secBons
Sequence the quesBons in each secBon Sequence the secBons
Blocks can be used in organizing ques9ons with a common
theme
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 21
Guidelines in construcBng quesBonnaires Sequencing /
Organiza%on
ConvenBons Include date and loca9on of interview, and name of
interviewer to facilitate quality control
Usually start with iden9ca9on sec9on / socio-demographic
sec9on
Start with easy ques9ons (i.e. neutral / not too personal
quesBons rst) For very long quesBonnaires, ask easy ques9ons rst,
more dicult ques9ons in the middle, and then go back to easier
ques9ons
Guidelines in construcBng quesBonnaires Sequencing /
Organiza%on
ConvenBons Ask general ques9ons rst before specic ones Earlier
quesBons should not inuence response to subsequent ques9ons
Include quesBons to cross check responses
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 22
Guidelines in construcBng quesBonnaires Sequencing /
Organiza%on
ConvenBons Use introduc9on or transi9on statements before
starBng another block / secBon / group of quesBons ExplanaBon of
what to expect in the next block ExplanaBon of the importance of
ques9ons in the next block
ReiteraBon of conden9ality for sensi9ve ques9ons
Guidelines in construcBng quesBonnaires Sequencing /
Organiza%on
ConvenBons For aotude quesBons, combine posi9vely- and
nega9vely-stated statements Increases variety Decreases response
set
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 23
Guidelines in construcBng quesBonnaires Sequencing /
Organiza%on
ConvenBons Follow natural chronological order for some variables
Ask educaBon rst before occupaBon Ask whether one smokes rst before
asking the number of sBcks usually smoked
Guidelines in construcBng quesBonnaires Forma_ng
Start with the iden9ca9on page Use introductory statements May
parBBon the page into two columns one column for quesBons, and the
other column for answers
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 24
Guidelines in construcBng quesBonnaires Forma_ng
Recording of responses Use check boxes or numbers to be
encircled for pre-coded or open-ended quesBons
Provide ample space for answers to open-ended quesBons
Provide skipping instruc9ons and other instrucBons for
interviewers
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 25
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 26
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 27
Phrasing the quesBons Clear
Entails awareness of the level of understanding or educaBon of
respondents
Do not use technical terms or complex words or phrases. Use
short, simple, and direct ques9ons Do not use ambiguous ques9ons Do
not use words that are prone to mis-pronouncia9ons
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 28
Phrasing the quesBons Clear
Avoid double-barreled ques9ons When two quesBons are implied in
what is meant to be one.
Example: Do you agree or disagree that AIDS can be transmi7ed
through hand shaking or through other forms of physical
contact?
Phrasing the quesBons Unbiased
The quesBons should not inuence the way the respondents will
answer
QuesBons must not reveal the person / group of the study if it
can lead the respondent to answer in a parBcular way.
QuesBons should not favor one side of the idea. QuesBons should
not lead the respondent to give answers when they are not qualied
to do so.
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 29
Phrasing the quesBons Tacbul
QuesBons should not embarrass the respondent especially: If the
topic is sensiBve If quesBons assume that the respondent has
knowledge on the issue when he / she has none
Phrasing the quesBons Adequate
The quesBonnaire should include all the necessary informa9on to
allow accurate administraBon, and recording of responses Necessary
explanatory material are present Pre-coded responses are mutually
exclusive and exhausBve
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 30
Phrasing the quesBons Neither loaded nor leading
Loaded ques9on: emoBonally charged or has a false or disputed
presupposiBon Example: Why did you engage in pre-marital sex?
Leading ques9on: gives the respondent an idea of desired
response Example: Because AIDS is transmissible, do you think
pa%ents living with AIDS should inform their sexual partners rst of
the infec%on status?
Open-ended quesBons
The respondents are free to give whatever answer/s he / she nds
relevant
The respondents can phrase answer in his / her own words /
responses
Examples In your opinion, how is leptospirosis transmi7ed? What
is your primary considera%on when purchasing food items for your
children?
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 31
Open-ended quesBons
ADVANTAGES
SBmulate free thought, solicit sugges9ons, probe peoples account
of events, clarify posi9ons
Used in exploratory / qualita9ve studies
Can be used in pilot studies to guide the development of
close-ended quesBons
DISADVANTAGES Entails much recall and organiza9on of thought
Probing is necessary Unsuitable for SAQ Dicult to record
responses Dicult to process and analyze
Longer interview 9me
Close-ended quesBons
Answers or choices are already provided to the respondents.
May involve raBng or ranking of responses May involve asking
whether the respondent agrees or disagrees with a statement
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 32
Close-ended quesBons
ADVANTAGES
Easy to analyze Can easily measure levels or degrees of a given
variable
DISADVANTAGES
Choices are limited to pre-coded responses thus there is a risk
of missing important dimensions of what the respondent knows,
believes in, or feels about
RaBng scales
Can be used as an alternaBve to close-ended quesBons to elicit a
graded response.
Example: On a scale of 0 to 10, where 0 indicates complete
disagreement and 10 indicates complete agreement, please rate you
degree of agreement or disagreement on mandatory HIV tes%ng to all
incoming medical students?
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 33
Pre-tesBng
Conduct of the data collec9on process to 20-50 respondents that
have characteris9cs similar to the study popula9on (i.e. for
knowledge and aotude scales: 5 respondents per quesBon)
Things to look for High number of dont know responses or items
les blank Incomplete quesBonnaires Inconsistent answers Most
responses fall under one category Responses are irrelevant to the
study
References
Borja M (2012). Lecture slides on data collecBon. UP CPH
DEBS
Daniel W & Cross C (2013). BiostaBsBcs: A FoundaBon for
Analysis in the Health Sciences. John Wiley & Sons, Inc.
Kumar R (2011). Research Methodology: A Step-by-Step Guide for
Beginners. Sage: London
Mendoza O et al (2010). FoundaBons of StaBsBcal Analysis for the
Health Sciences. Department of Epidemiology & BiostaBsBcs,
College of Public Health, University of the Philippines Manila:
Manila
-
3/26/15
Amiel Nazer C. Bermudez, MD, MPH 34
Thank You J