5060 Flashcards
research hypothesis
statement that claims a relationship exists between the independent and dependent variable
usually what the researcher is trying to prove
null hypothesis
statement that claims that NO relationship exists between the independent and dependent variable
independent variable
manipulated variable
usually the intervention being researched
dependent variable
affected by the independent variable
extraneous variables
variables that are not being studied that can affect outcome of research
experimental group
group that is being experimented on
control group
group that is not experimented on
provides a baseline for comparison
control
measures that are taken to reduce the influence of extraneous variables on the dataset
examples of control
homogenous sample
consistent data-collection procedures
manipulation of independent variable
randomization
types of research design
experimental
quasi-experimental (no randomization)
non-experimental
what type of research design is most common in nursing?
quasi-experimental
types of experiment designs
true or classic
solomon four group
after only
nonequivalent control group
after-only nonequivalent control group
one group pretest-posttest
time series
types of nonexperimental designs
correlational studies
developmental (cross-sectional, longitudinal/prospective, retrospective ex facto)
correlational study
examine relationship between 2+ variables
cross-sectional study
outcome among individuals at one point in time
longituindal/prospective
changes in individuals over time
retrospective/ex post facto
variable x is related to variable y, but x can’t be randomized or measured
find a group without x and compare to a group with x and see if there is a difference in y
threats to internal validity
history
maturation
testing
instrumentation
mortality
selection bias
threats to external validity
selection
reactive
measurement
3 characteristics of a good research question
well-defined population
well-defined variables
testability
3 types of research questions
correlational
comparative
experimental
directional hypothesis
hypothesis states a relationship exists and in what way the data will trend
non-directional hypothesis
hypothesis states a relationship exists but does not predict how the data will trend
descriptive statistics
summarize or describe features of the data set
re: central tendency or dispersion
not the actual results of the data itself
usually displayed in visuals (table, graph, histogram)
inferential statistics
make generalizations about the population based on sample statistics and hypothesis testing
p value
probability
represents how likely x is to happen
evaluates how well the data supports the null hypothesis
high p value
data supports null hypothesis
>0.5
low p value
data does not support null hypothesis
<0.5
confidence interval
value that sample is believed to lie within
normally measured at 95%
CI <95% considered not statistically significant
correlation coefficient
measures strength and direction of relationship between 2 variables
range of values for correlation coefficient
-1.0 - 1.0
-1.0 = strong negative (inverse) relationship
+1.0 = strong positive relationship
0.0 = no relationship
measures of central tendency
mode
mean
median
mode
number that occurs most frequently in a set
unstable
median
middle number
not very useful
mean
arithmetic average of all numbers
affected by outliers
standard of deviation
measure of how data is spread around the mean
types of skew
positive (right-side occurs w/ lower limit)
negative (left-side, occurs w/ upperlimit)
measures of variability
range
standard of deviation
percentile
percentile
percentage of scores that a given score exceeds
standard of deviation
average deviation of scores from mean
how far each value lies from the mean
high = far from mean
low = clustered around mean
semi quartile range
range of middle 50% of scores
half of the difference between the upper quartile and lower quartile
level of measurement
nominal
ordinal
interval
ratio
nominal
classification
numbers don’t carry any hierarchal value
info organized into descriptors (red hair, blue eyes, etc)
ordinal
relative ranking
ranks things against each other but differences may not be equal
(likert scale, income, education)
interval
items on scale are ranked in order with equal intervals between values but there is no absolute zero
(temperature, pH, test scores)
ratio
items on scale are ranked in order with equal intervals between values and an absolute zero
(income, weight, height)
normal distribution or empirical rule
1sd = 68%
2sd = 95%
3sd = 99.7%
types of statistical tests
t test
ANOVA
pearson R
chi-square
fisher’s exact probability test
non-probability sampling
convenience
quota
purposive
probability sampling
simple random
stratified random
cluster
systematic
qualities of rigour
credibility
auditability
fittingness
credibility
truth of findings as judged by others or experts in the field
auditability
accountability as judged by adequacy/comprehensiveness
all info should be provided
fittingness
faithfulness to everyday results of participants
reliability
does a tool accurately measure what it is supposed to
ratio of accuracy to inaccuracy
reliability tests
test-retest
parallel or alternate form
split half
item to total
interrater
kuder richardson
cronbachs’ alpha
validity
does an instrument measure what it is supposed to
validity tests
content validity
criterion related validity
construct validity
levels of measurement
nominal
ordinal
interval
ratio
data collection procedures
physiological measurement
observational method
questionnaires/surveys
interviews
records
available data
hypothesis testing
testing whether a scientific hypothesis is true or not by rejecting or accepting the null hypothesis
cannot PROVE the scientific hypothesis only find evidence to support/reject it
type 1 error
wrongly reject the null hypothesis when it is true
false positive –> wrongly conclude there is an effect when there isn’t
type 2 erro
wrongly accept the null hypothesis when it is false
false negative –> wrongly conclude there is no effect when there is
level of significance
probability of making a type 1 error
a = 0.05
interval measure tests
t test
anova
nominal or ordinal measure tests
chi square
sign test
signed rank
mann-whitney u
how to pick a statistical test
nature of research question
number of groups (2 vs. >2)
level of measurement
underlying distribution
p-value
probability of an event occurring. evaluates how well the data supports the null hypothesis.
p value < a (0.05)
reject null hypothesis
accept scientific hypothesis
p value > a (0.05)
accept null hypothesis
reject scientific hypothesis
level of significance
alpha usually set at 0.05 or 5%
maximum acceptable probability of making a type 1 error
t test
parametric
examine causality
requires interval/ratio LOM
tests for sig differences between 2 samples
most commonly used test of differences
anova
parametric
tests for differences between means (usually >2)
more flexible than other tests
ratio LOM
pearsons correlation
parametric
estimate degree of association between 2+ variables
test of CORRELATION
interval/ratio LOM
chi square
non-parametric
used with nominal data
test for differences between frequencies expected if groups are alike
indicates significant difference
how to prevent making a type 1 error
set the level of significance (a value) lower
if p is less than the a value the researcher rejects he null hypothesis & concludes results are statistically significant
how to prevent making a type 2 error
increase the sample size
threats to internal validity
history
selection bias
mortality
maturation
testing
instrumentation
