Experimental Research Methods Flashcards
grouped frequencies?
every value between and including first and last value
eg. 0-9, 10-19, 20-29
stratified random:
groups first formed to ensure even representation of groups
M&F/race
relative frequency?
percentage of all elements that fall within each class interval
frequency / sample size
cumulative frequency?
percentage of distribution that lies in or below that particular group
mode:
most frequent observation
median:
mid point of the distribution (50th percentile)
mean
arithmetic average
what is a deviation score?
element - mean
how do you calculate variance?
mean of the SQUARE of all deviation scores
variance = sum (X - mean)^2 / number of elements in population
what is different about sample variance?
divide by sample size - 1
instead of just normal sample size
what is the standard deviation?
the square root of the variance
what % of distribution falls within 1 standard deviation?
68%
what % of distribution falls within 2 standard deviations?
95%
what % of distribution falls within 3 standard deviations?
99.7%
what % of distribution falls ABOVE or BELOW 2 standard deviations?
95% within 2 standard deviations
therefore 5% above and below
or 2.5% below, and 2.5% above
what % of distribution falls below 1 standard deviation?
well 68% within 1 standard deviation
and 16% below -1 standard deviation
68+16 = 84
what is the standard deviation a measure of?
measure of variability
how do you calculate SEM - Standard Error of the Mean
standard deviation / square root of n
what is the null hypothesis?
this hypothesis you assume to be true
Ho = whatever you’re researching has NO effect -> no difference in mean
what is the alternate hypothesis?
Ha or H1 = whatever you’re researching does have an effect -> difference in mean
at what value do you reject the null?
if you get a p value of < 0.05
What is the P-value?
- “probability” value
- % under the curve converted to decimal
eg. 0.3% under curve(p-value), 0.002 probability value
what is a type 1 error?
error of rejecting the null hypothesis even through it is true
what is a type 2 error?
failing to reject the null hypothesis, when the alternate hypothesis is actually true
what is degrees of freedom and how is it calculated?
expression of sample size
df = n - 1
how do you calculate chi-square?
sum of (observed - expected)^2 / expected
aims of medical research?
- cure disease
- prevent
- improve productive lifespan
- save $$ from health cost
what are the governing principles for the Australian code for the care and use of animals for scientific purpose?
3 R’s
Replacement
Reduction
Refinement
describe “replacement” principle:
- ethical animal conduct
Replacement = methods that permit a given purpose of an activity or project to be achieved without the use of animals
eg. use of human instead - they can give consent for the most part
describe “reduction” principle:
- ethical animal conduct
methods for obtaining comparable levels of info from the use of fewer animals in scientific procedures or for obtaining more into from the same number of animals
eg. use pre-existing data
describe “refinement” principle:
- ethical animal conduct
methods that alleviate or minimise potential pain and distress, and enhance animal wellbeing
eg. allow animals to “acclimatise” to new environment and equiptment
levels of randomised control bias:
1 - systematic reviews of RCTs
2 - one RCT
3 - pseudo-randomised controlled trial
3.2 - comparative study with concurrent controls
3.3 - comparative study without concurrent controls
4 - case-series
describe the structure of a scientific study write up?
abstract intro methods results discussion references
describe prediction by interpolation:
when dosage lies within tested range
describe prediction by extrapolation:
new dosage outside original range
steps in experimental process?
1) research question
2) hypothesis
3) data analysis
4) conclusion
how can we typically calculate the 95% Confidence Interval?
SEM x z-score
z-score for 95% confidence = 1.96
what can be assumed about the 95% Confidence Interval if the P<0.05?
the 95% confidence interval will not contain a zero
what does the P-value allow you to answer?
if there is no correlation between the 2 variables, what is the chance of observing a correlation as far from zero as we have observed, by chance(random sampling) alone?
what does placebo stop people from doing?
stops them hunting around for another treatment because many may believe they have the treatment
pseudo-random
kinda random but not, you can figure out the pattern
like dates of birth, days of the week
what is the null hypothesis:
whatever you’re researching has NO effect
-> no diff. in mean
steps of hypothesis testing?
1) state the Ho and Ha
2) select decision criteria (P<0.05)
3) establish critical values (direction & df)
4) draw random sample, calculate mean
5) calculate S & SEM
6) calculate t-statistic for sample mean
7) compare t with critical values -> reject or retain the null hypothesis
what could increase the power of an experiment?
- eliminating inter-subject variation (eg. age)
- increased sample size
- minimising instrument errors
define knowledge translation
the synthesis, exchange & ethically sound application of knowledge by relevant stakeholders to accelerate the benefits of global and local innovation in strengthening health systems and advancing peoples health
what is knowledge creation?
primary research
what is knowledge distillation?
creation of systematic reviews/guidelines
what is knowledge dissemination?
appearances in journals and presentations
what people are involved in knowledge translation?
health care professionals/providers(clinicians)
patients
managers
policy makers
identify and briefly discuss 4 factors that are known to influence how easily research evidence is translated into policy or practice:
1) what: is being implemented, characteristics of implementation
2) who: characteristics of target groups,
3) where: features of the context in which you are attempting to implement
4) how: process used to facilitate implementation
eg. generally need tailored interactive and multi-faceted approach
what 3 assumptions underpin the knowledge translation metaphor?
1st: ‘knowledge’ equates with objective, impersonal research findings
2nd: it is useful to conceptualise a ‘know-do’ gap between scientific facts and practice
3rd: practice consists of a series of rational decisions on which scientific research findings can be brought to bear
why does the ‘know-do’ gap exist?
1) lack of standardised and up to date practice guidance
2) need for a better design and more effective use of electronic information system
in terms of qualitative studies what could be some patient level factors?
- perceived benefits and drawbacks
- ability to self-manage(if applicable)
- acceptability of equipment used,
- desire/demand for extra support/technology/
in terms of qualitative studies what could be some clinical level factors?
- personal experience
- politics and personalities
- trust in evidence base
in terms of qualitative studies what could be some organisation level factors?
- service investment requirements
- impact on non-pump population
in terms of qualitative studies what could be some system level factors?
- commissioning pump services
- who pays ad how
language and terminology likely to be used to describe the process of moving research into policy and practice:
knowledge translation, mobilisation, utilisation, exchange, transfer
implementation science, translational science, research utilisation
define efficacy:
ability of an intervention to produce results under ideal conditions
eg. RCT with well staffed & motivated patients
define effectiveness:
ability of intervention to produce results in general use
define mobilisation:
moving available knowledge into active use
define utilisation:
extent to which potential access is converted into realised access
what is implementation science?
study of methods to promote the integration of research findings and evidence into healthcare policy and practice
how can qualitative research help with decision making?
1) directing and informing policy and program development
2) assisting in program design
3) considering individual health decisions, actions, impacts
4) directing and informing policy and program evaluation
3 considerations - in place of experiments in the universe of health research:
1) efficacy and effectiveness
2) questions of meaning
3) public health approaches
what is PICO?
a framework for evidence-based decision process - study design
helps you answer clinical questions using research data
what does PICO stand for?
P = patient/populaiton I = intervention C = comparison/control O = outcomes
define a systematic review:
a review in which there is..
- comprehensive search for relevant studies on a specific topic
- studies identified are then appraised
- and synthesised according to a pre-determined and explicit method
meta-analysis:
- quantitative analysis of results
- putting available studies together increases statistical power
=> decreases likelihood of type 1&2 errors - generalisability of results improved by combing studies
when do you use a one sample z-test?
when there is just 1 mean and you know the standard deviation
when do you use a one sample t-test?
when there is just one mean but you DO NOT know the standard deviation
when do you use a 2 sample t-test?
when there are 2 means
matched/paired/pooled t-test?
used when 2 measurements on each subject
when do you use chi-square test?
OBSERVED
EXPECTED values
what is regression/correlation useful/often used for?
to predict the value of one variable based on another variable
what is the equation of a line
y = mx + b
Y = b0 + b1*X
b0 = intercept b1 = slope
how do you calculate slope?
y = y2 - y1 / x2 - x1
requirements of a well designed study
1) absence of bias
2) sufficient power
3) exploration of range of applicability
4) keep it simple: minimise chance of making mistake
5) amenable to statistical analysis
why do experiments with/in humans?
- results applicable “human are humans”
- can directly study the actual disease/problem(model not the same)
- subjects can describe how they’re feeling
- humans can consent to participate
- test real life situations
- can be cheaper
why do animal experiments?
- test causality or establish mechanisms
- eliminate variation(genetic, environmental), confunding
- some experiments are impossible in humanseg.killing to collect tissue
- establishing safety before human
- no problems with adherence to protocol
- get enough subject
what different categories of people does an animal ethics committee consist of?
A: vet
B: recent/current experience in use of animals for scientific purposes
C: RSPCA/animal welfare
D: random person, generally older person
E: person looking after animals within institution
when using animals you need to consider?
3 R’s
Replacement: achieved without use of animals
Refinement: minimise pain, enhance wellbeing
Reduction: fewer animals, or same info from same amount of animals
what are the NHMRC levels of evidence for interventions?
I: systematic review trial
II: one RCT
III: pseudo-RCT
III-2: comparative study with concurrent controls
III-3: comparative study withOUT concurrent controls
IV: case-series
what is randomisation?
allocation of treatment to trial participants based on chance
what is selection bias?
good generation of random sequence that cannot be determined by anyone(concealment of allocation)
*before trial starts
what is performance bias?
blinding of participants and personnel
*once trial has started
what is attrition bias?
maintaining blinding of participants especially
*avoid placebo group sourcing treatments else where
what is detection bias?
blinding of outcome assessors especially
- avoid favouring of certain participants
what is reporting bias?
select the good results to report, or conflict of interest
what are the 4 pillars of ethical research?
1) merit & integrity
2) justice
3) beneficience
4) respect to persons
describe pillar 1: merit and integrity of ethical research:
- honesty in research
- ensuring legitimate search of knowledge
describe pillar 2: justice of ethical research:
- do not “lie”
- selection, recruitment, exclusion and inclusion of research participants must be reasonable
- fair distribution of benefits and burdens
describe pillar 3: beneficence of ethical research:
- minimise risk of harm to participants
- offers benefits to society
describe pillar 4: respect to persons of ethical research:
recognition of their intrinsic value, respect privacy, confidentiality
what is the most important pillar of ethical research?
respect to persons!
describe informed consent:
voluntary, based on sufficient info, adequate understanding of research and implications of participation
*basically no coerce - cannot black mail/pressure/manipulate people to do shit
what does it mean if the confidence interval line on a forest plot crosses the mid line?
if the confidence interval line of a study crosses the mid line the results of the study are NOT statistically significant
in a forest plot what does the size of each diamond show?
the size of each data point or diamond shows sample size of each study
when studies are put into funnel plot what does it show?
it shows presence of performance bias within the study
in a funnel plot what is along the x axis?
favours treatment —————————— favors control
in a funnel plot how would you identify the presence of bias?
when there is a big gap
- has been in right bottom in all diagrams i’ve seen
name the major types of variables and their sub-topics:
categorical= regular(M/F, nationality) & ordinal(categories ascending & descending) numerical= continuous -> interval, ratio & discrete -> counts(# of kids)
example of catagorical regular data:
M/F
marital status
nationality
example of categorical ordinal:
things that can be placed in ascending or descending order
what type of data would age be?
numerical continuous ratio
what type data is size?
categorical ordinal
what type of data is sex?
categorical regular
what type of data would the number of kids be?
numerical discrete
what hypothesis do you assume to be true?
null
2 possible consequences for not blinding?
1) performance bias - blinding of patients & personnel
2) attribution bias - maintaining patient blinding to prevent drop out and seeking treatment else where
if by chance the control group of a study contained bias eg. more horrible children or older average age how would this influence interpretation of results?
..what statistical test could analysis this accurately?
control would have higher or lower dependent variable at baseline
repeated measures analysis
-> looks at change from baseline in each group
in a question where 2 measures are taken from the same subject what statistical test is likely to be used?
paired t-test
in a question where there are 2 means what statistical test is likely to be used ?
2 sample t test
in a question where there are 2 cohorts with no association to each other in a study what statistically test is likely to be used?
chi-squared
what is a major purpose of random assignment in a clinical trial?
to help ensure that study subjects are representative of the general population
how do you find the mean if the study is paired?
paired means 2 measures are taken on the same subjects
therefore you must find the mean difference for each participant
eg. post-treatment x - pre-treatment x
then sum the differences and divide by sample size = sample mean
what is the difference between df with 2 means?
(n treatment + n control) - 2
if people are recruited from different places how does this effect the results of the trial?
it shouldn’t effect the results of the trial as long as participants are properly randomly allocated
(random allocation prevents selection bias)
what kind of bias is it when people seek other treatment?
performance bias
what kind of bias is it when participants have a lack of interest?
attribution bias
how would you go about answering this question?
“Is attrition bias likely to be a problem in this study?”
think of all the reasons why participants could become demotivated and drop out of the study