Flashcards
quiz2
Cards in this deck (60)
- 1Operationalizationthe process of understanding what is measured,
how the indicators are measured, and the rules used to assign a value
to what is observed and to interpret the value
HOW YOU WILL MEASURE a certain characteristic of
human behavior or condition, to turn it into a variable - 2Measurement rulesrefer to steps for assigning labels in an explicit
and consistent manner i.e. count ever occurrence of a certain
behavior in a given period of time. - 3Direct relationshipDependent Variable varies as a linear function of Independent Variable.
- 4Mediated relationshipthe effect of IV on the DV is explained (mediated) by the
third variable (mediator) i.e. MV is influenced by the IV; and MV influences the DV - 5Moderated relationshipthe magnitude of the effect of the IV on the DV varies as a
function of the third variable (moderator) - 6Spurious (confounding) relationshiptwo variables in a dataset are
mathematically related, but there’s no apparent mechanism to support
causal link between them. - 7Spurious (confounding) relationshiptwo variables in a data set are
mathematically related, but their relatedness is due to a third
variable (confound) that, in fact, impacts both IV and DV - 8Unit of analysisthe smallest observable unit used in a particular study
- 9Nominal level measuresProvide information about whether the observation is qualitatively
different from another i.e. gives a name to your data/observations.
– Examples: race, ethnicity, marital status, political affiliations,
educational level, social agency type, geographical location.
■ Categories are mutually exclusive.
■ Assign values to categories [i.e. 1 =single, 2 =married, 3 =divorced].
■ However, these absolute values are meaningless and have no
quantitative value or mathematical interpretation. - 10Ordinal level measuresProvide information about the relative amount of one observation
compared to the other (less than, greater than, etc.).
■ Ability to rank or position the categories of the variable in some sort of
line i.e. 1 comes before 2 and 3 comes after 2; and the values exist in a
rank from lowest to highest, poor to excellent, etc.
■ Precision i.e. exactness or amount of detail needed e.g. how serious, how
often, how sure, how important, etc.
■ Consider: 1= Agree, 2 = Not Sure, 3 =Disagree; and 1= Absolutely
Disagree, 2 = Disagree, 3 =Not sure, 4 =Agree, 5 =Absolutely Agree - 11Scale/Interval/Ratio level measuresScale data are real numbers i.e. they have actual quantitative meaning
- 12Interval level- ability to know the value of space between measurement
points [equal intervals or fixed measurement]
– Example: the distance between 10-20 is the same as the distance
between 30-40.
– No absolute value i.e. temperature, intelligence scores i.e. a zero value
does not indicate complete absence of a measured variable - 13Ratio levelSame as interval data but has an absolute zero i.e. a zero
value is meaningful number. e.g. age in years/months, income in dollars,
height, weight, time, test scores, etc.
– Ratio data/numbers can be transformed i.e. added, subtracted, multiplied and
divided. - 14Ease of usewhether the client or practitioner will be able and willing
to use a given measure. - 15Relevance to intervention planninghow useful it will be to the
success of the intervention - 16Accessibilitysome measures are difficult to obtain i.e. expensive, not
available for use by other professionals, without the appropriate
training. - 17Characteristics of the clientage, literacy level, SES, language
- 18Directnessdirect versus indirect measures
- 19Sensitivityability to detect change [improvements or deteriorations]
in the event or behavior. - 20Random errorobserved values tend to be inconsistent when the same
characteristic/behavior is measured repeatedly under the same conditions.
Possibility that observations could be affected by chance. - 21Systematic error:tendency to err in a particular way (bias).
– Biases lead to consistent overestimating or consistent underestimating of
the same characteristics being investigated. - 22ReliabilityConsistency of measurement.
– Reliability of the measurement method is its ability to minimize the impact of
random error on the observed score. - 23ValidityThe validity of a measure indicates whether and how well it measures what
is intended i.e. the method’s ability to capture the True Value of one’s
construct of interest.
■ Validity of the measurement method is its ability to minimize the impact of
both (a) random error and (2) systematic error on the observed score. - 24ReliabilityConsistency of a measurement i.e. the extent to which a
measure produces the same results when used repeatedly to
measure the same behavior/event.
■ Test of reliability requires that observations be independent and be
made under the same circumstances.
■ Reliability values are expressed as correlations [between 0 and 1] or
percentage of agreements.
■ The higher the reliability values, the better. Preferably .80 or 80% and
above. - 25Test-Retest Reliability AssessmentDetermined by independently measuring the same group of
people/things, with the same measure on different occasions under
the same circumstances.
■ Difference between observed (measured) scores at different time
points represents random error.
■ Less random error signifies higher reliability of the measurement
method
■ Test-retest reliability can be calculated for any kind of measurement
method, including self-report instruments, and those involving a
measurement “rater”, such as interviewer. - 26Internal Consistency ReliabilityOnly for assessing reliability of Multi-item scales i.e.
– How consistently multiple items (questions) of the scale measure the facets of
the same construct.
– How well do item responses correlate with each other?
■ How to assess internal consistency:
– Split-halt method – determined by dividing items in a measure into two
groups with equal number of items, computing a score for each half of the
measure and determining the correlation between scores on the two halves.
– Parallel forms methods – for scales/exams/tests that have multiple
versions: correlate scores of the same respondents on different versions of
the same test
– Coefficient alpha also known as Cronbach’s alpha – statistic used to
quantify internal -consistency reliability i.e. average interrelation among a
set of items, weighted by the number of items - 27Inter-observer/Inter-rater Reliability AssessmentThe extent to which two or more observers are consistent with each
other when they independently observe or judge the same event or
behavior.
■ Quantified using either a correlation coefficient or percentage of
agreement i.e. percentage of times both observers agree with each
other in their observations.
■ Scores derived by several interviewers are compared. Difference in
scores represents random error. Less random error signifies higher
inter-rater reliability of the measurement method - 28Internal validityindicates whether and how well it
measures what is intended. - 29External validitypertains to the generalizability of the treatment
effects to other populations, settings, treatment or measurement
variables. - 30Criterion-Related Validityextent to which a measure is related to an outcome
- 31Concurrent Validity assessmentUse your new depression scale at the same time with another, reputable
measure of depression (e.g. Beck’s scale), with the same individuals. Correlate
their depression scores on your scale with their scores on Beck’s scale. If
correlation is high, your scale has high concurrent validity.
– Another way to assess Concurrent Validity: Known-groups validity: administer
your new depression scale to a group of people known to have depression (e.g.
from clinical assessment). - 32Predictive Validity assessmentSame thing as Concurrent Validity, except the criterion is observed later. If
depression scale data at Time1 strongly correlates with clinical depression
assessment at TIme2, your scale demonstrated good Predictive validity. (The
scale’s readings predicted future clinical status.) - 33Construct Validityis your scale measuring the RIGHT
CONSTRUCT? - 34Convergent / Divergent Validity assessment:To make sure your scale is not picking up information about a
different construct (your scale diverges from the different construct; or
does not unduly converge with it). - 35Face validityDo items in the scale seem to capture the actual facets of the
construct of interest? - 36Content validityDoes it seem that the items of the scale capture the full range of
meaning of the concept? - 37Administration proceduresFactors Influencing Reliability and
ValidityDistractions i.e. chaotic, uncomfortable, noisy environment
– Instructions provided to participants
– Conditions of measurement i.e. sensitivity of questions, confidentiality
concerns
– Systematic changes in the measurement conditions i.e. morning/evening - 38PopulationsFactors Influencing Reliability and
the social or cultural characteristics of the population i.e. race,
cultural background, gender, age, sexual orientation, SESValidity - 39Threats to ValidityHistory
– Did any other events occur that may be responsible for the particular
outcome other than the intervention?
■ Maturation
– Did any psychological or physiological change occur within the client that
might have affected the outcome?
■ Testing
– Taking a survey the first time may sensitize the client and influence their
subsequent scores
■ Drop out
– clients drop out of treatment - 40PopulationTotality of persons/observations/objects with which the research is
concerned. - 41SampleA proportion of persons/observations/objects drawn from a population of
interest. - 42Sampling FrameList of every person/observation/object in your population of interest
- 43Sampling errorNatural tendency of a sample to imperfectly reflect a population
- 44Parameterthe average value of a quantifiable characteristic in the population. Can be
thought of as the population’s TRUE VALUE of the given characteristic.
– Example: Parental Self-Efficacy (PSE) among immigrant populations tends to
be lower than PSE among non-immigrant populations. - 45Parameter Estimatean “educated guess” about the population parameter, based on measuring the
characteristic of interest in the SAMPLE representing the population.
– The better the sample represents the population, the closer the parameter
estimate will be to the actual parameter i.e. the relevant characteristics of the
sample must be close to those of the population (e.g. similar age range,
education, income, etc.). - 46Sampling erroran inevitable difference between
parameter estimate and the true population parameter.
Parameter Estimate = Parameter + Sampling Error - 47Margins of Errorthe range of values below and above the sample
statistic - 48Determining a Sample SizeResources e.g. budget, time and human resources
■ Research design – do you need a random sample?
■ Population size i.e. If the population is small, include all
■ Feasibility and data collection strategy
– large vs. small programs/agencies
– in person, telephone or mail survey
■ Minimum acceptable level of precision and confidence interval - 49Probability samplingEvery person/object in the study population has an equal chance/probability of
being selected to participate.
– Requires a sampling frame
– Expectation: the sample should be representative of the population.
– Ability to generalize the findings /external validity. - 50Non probability samplingNot all people in the population have the same chance of being included in the
sample.
– Usually conducted in Exploratory studies where the purpose is just to collect as
much data as possible.
– Does not require a sampling frame - 51Simple Random SampleA simple random sample guarantees that every member of the
population has an equal chance to be selected.
§ Define your study population i.e. Who/what to include –sampling
frame
§ Label your study population and decide on your sample size (n)
– Assign a number to each member in your population sample
§ Label should have the same number of digits as your sample size (n)
– For a sample size <10, use single digits: 1, 2, 3, etc.
– For a sample size of 10-99, use 2 digits: 01, 02, 03, etc.
– For a sample size of 100-999, use 3 digits: 001, 002, 003, etc.
■ Then randomize on this number –using a table of random numbers - 52Systematic Random SampleSample size is expressed as a ratio of the population
From your sampling frame, determine your sample size in terms of
percentages of your population (e.g. 10%, 20%, 25%, 50%, etc.)
■ Select your sample using an interval based on your percentage. - 53Stratified random samplesSample of subgroups within the population (strata), usually underrepresented
subpopulations.
– Guarantees that each subgroup will be represented
– Each stratum may be proportionate or disproportionate to the sample size i.e.
divide population into strata (different subgroups) and randomly select people
within each stratum. - 54Random cluster samplesUseful in creating difficult sampling frames i.e. when cases are geographically
separate or in “natural clusters”. i.e. schools, hospitals, city blocks, prisons etc.
– Randomly select clusters and include all individuals within a cluster. - 55Convenience/ Availability SampleSelecting the sample from the first cases/people who are
available/accessible. For example, group of friends, students, etc. - 56Purposive/Purposeful sampleUsed when we want to purposively choose a particular group i.e. with unique
characteristics - 57Quota sampleAttempt to be inclusive – based on theory
– Same as stratified sample bust individuals are not selected randomy.
– Usually result into some groups being over sampled and others under
sampled. - 58Snowball sampleSample from referrals and recommendations from respondents.
– Useful to recruit individuals in hard-to-reach groups. - 59Typical case studiesCases that represent the average
- 60Maximum Variation, Deviant and Extreme case samplingExtremes at either end of range (max/min) of client's experiences i.e. Cases
that do not fit the norm.
Information
- Description
- Flashcards on quiz2, created by Melanie Hampton on 05/03/2018.
- Created by
- Melanie Hampton
- Views
- 0
- Rating