Exam 2 Quantitative Analysis Flashcards
Statistical Analysis
Purpose - findings that are significant, also clinically significant
Probability Sampling
What is it?
Sampling method in which elements in the accessible population have an equal chance of being selected for inclusion in the study
Probability Sampling
What is a biased sample?
Doesn’t look like the population ‘you’ are wanting to look at
Probability sampling
What is sampling error?
What ‘we’ got, vs the truth
Probability sampling
how does sample size affect statistics?
Larger the sample, better off to find relationships…if you’re looking for subtleties, need larger groups
Types of stats
Descriptive
describe sample, what sample looks like, basic relationships
Types of stats
inferential
helps make generalizations
Types of stats
Univariate
One variable (ex.age)
Types of stats
Bivariate
two variables
Types of stats
multivariate
> two variables
Types of stats
Two categories of inferential
Parametric
Nonparametric
Descriptive
Describing the sample
Descriptive
Frequency
Stem-and-leaf plots (hash marks)
histograms (Bar graph)
frequency polygons (line graph)
Central tendency
Mode
Most frequent score
Central tendency
Median
Middle score
Central tendency
Mean
Average score
Normal curve
Symmetrical. “Bell curve”
Normal curve is very rare when dealing with people…going to have outliers
Skewed curve
Asymmetrical
Tail is right sided…positive
tail is left sided…negative
Kurtosis
How tall or flat the curveis
Range
spread of scores
difference between highest and lowest score
Semiquartile or interquartile range
the range of the middle 50%…divide range into quarters, discard Q1 and Q4…gets rid of the outliers
Standard deviation
How far someone is away from the average
Bivariate correlations
Two groups
looking for RELATIONSHIP between two scores (similarities)
Direction of relationship - negative
X goes up, Y goes down
Direction of relationship - positive
X goes up, Y goes up
Magnitude or strength of relationship
Weak…score close to 0
Strong…score close to +1 or -1
Remember… -0.6 is stronger than +0.2
When use pearson r
if using interval or ratio
When use spearman rho?
If using ordinal
Scatter plots
If the dots form a PERFECT line, regardless of direction, then perfect correclation
If the dots are completely scattered no semblance of a line, then no correlation.
If dots are in the general form of a line, but not perfectly straight…that is typical correlation.
Inferential statistics
Making suppositions about the population from information you know about the sample
Key points
Population = Parameter .... p - p Sample = Statistic.... s - s
What do Greek symbols indicate?
Inferring a population parameter
Significance level
setting alpha
.1, .05, .01, or .001
alpha is the risk we are willing to take that we will make a mistake and WRONGLY REJECT the null hypothesis when it is, in fact, true….ex….we should have accepted the null hypothesis, there is not relationship
Chance error
mastate results from a fluke or a chance occurrence that did not reflect reality
When is alpha set?
BEFORE collecting the data
Significance Level
Calculate our statistic
THEN check a table for the actual probability (p value) the result was due to chance
Significance level
when is null hypothesis rejected/accepted?
If p is less than alpha, reject null hypothesis
if p is greater than alpha, do not reject null
Deciding on alpha
Type I error…when researcher rejects null when should have been accepted (alpha)
Type II error…when researcher says there is no relationship (accepts null hypothesis) when there is actually a relationship (power)
all based on null hypothesis
Confidence intervals for clinical meaningfulness
An interval or range of scores in which the true value lies with a given degree of certainty
Confidence intervals for clinical meaningfulness
What does the interval width indicate?
A wide interval indicates that our data isn’t good enough to give us good confidence in our inferences about the population.
Wide interval = less confidence
Narrow interval = more confidence
If CI contains 0, there is no effect
If CI does not contain 0 , there is an effect
Types of inferential stats
Non-parametric
Inferential statistic involving nominal or ordinal level data to infer to the population
Types of inferential stats
parametric
inferential statistical tests involving interval or ratio level data to infer to the population
Types of inferential stats
basic assumptions to be met
Probability sample (Sampling method in which elements in the accessible population have an equal chance of being selected for inclusion in the study)
Normal distribution (bell curve)
Interval or ratio level data
reduction of error
Nonparametric
When you cannot assume that your sample looks almost exactly like the population(small samples…odd bell curve)
Chi square
Looks at differences in frequency of independent groups.
uses NOMINAL or a combination of NOMINAL and ORDINAL data
tablecompares actual data to the data expected to occur
Parametric
When you CAN assume that the sample is enough like the population that you can make big leaps in faith with your statistical conclusions
t-tests
Comparing TWO groups
groups are independent (not related) or dependent (correlated or related)
t-tests
what are you looking for?
differences in means between groups…get a “t” statistic
t-tests
independent variable
Nominal
t-tests
dependent variable
interval or ratio
ANOVA
comparing MORE THAN TWO groups
looking for differences in means between the groups
Gives you an ‘F’ statistic
ANOVA
Independent variable
nominal
ANOVA
dependent variable
must be interval or ratio
Simple Regression
seeks to predict the value of Y based on the value of X
Simple regression
ONE independent and ONE dependent variable
Simple Regression
Dependent variable
MUST be interval or ratio
Simple regression
calculates a slope (b) and intercept (y)
Reported as the regression coefficient ‘R’…upper case R…lower case r is pearson
R squared is the explained variance
Multivariate stats
when you have more than two comparisons or groups
ANOVA types
ANCOVA - analysis of covariance
MANOVA - multiple analysis of variance
MANCOVA - multiple analysis of covariance
All examine differences in mean
Regression type
multivariate statistics
Tries to predict an unknown value from the values known
Multiple regression - an inferential statistical test used to describe the relationship of three or more variables
logistical regression -
hierarchical regression
canonical regression
Other multivariate stats
Looking at structure
Discriminate function
factor analysis- a test for construct validity that is a statistical approach to identify items that group together
Structural Equation Modeling
Biostatistics
Round in Randomized Control Trials and other studies that focus on treatment outcomes or illness predictions
Biostatistics
What do you want to look at?
Magnitude of effect
Strength of association
Magnitude of effect
The effect is the rate of occurrence (usually seen as a bad thing…falls, pressure ulcers) in each of the groups on the outcome of interest
The DEGREE of the difference or lack of difference between the groups in the study
Strength of association
ABSOLUTE risk for EVERYONE getting…what you’re testing (example…breast cancer)
probability that a negative outcome will occur
Absolute risk reduction
when the risk for the negative outcome is less for the experiment group than the control group
example: experimental group…low fat diet…risk for breast cancer reduced
Absolute risk increase
when the risk for the negative is more for the experimental group than the control group
example…birth control study…found risk for breast cancer increased
Other strength of association
relative risk
likelihood that the negative outcome will occur in one group compared to another
Other strength of association
relative risk reduction
proportion of risk that the negative outcome will occur in the experimental group compared to control group
Other strength of association
Odds Ratio
***Used heavily in health care
Number exposed who actually get the negative outcome divided by number exposed who do not get the outcome
Other strength of association
Odds Ratio values
> 3 = strong, 1.3-3.0 = moderate, 1.1-1.3 = small, 1 = even, or 50:50 ods
Other strength of association
Number needed to treat
Number one would need to treat to prevent one case of the negative outcome from occurring
Other strength of association
Number needed to harm
Number one would need to treat before one case would be harmed
EBP Clinical Significance
- Think on an individual level
* What do the numbers mean when you are talking about real people?
EBP Clinical Significance Example: One person has a risk of stroke of 1% Another has a risk of stroke of 10% An antihypertisive drug has a relative risk reduction of 22% but has absolute risk increase in gastric bleeding side effects in %5 of cases Which person should take the medication?
The patient who has a 10% risk of stroke. The person with 1% chance…the 5% chance of GI bleed may not be worth it.
Using a computer
Complicated statistics
programs:
Excel
Access
SPSS
SAS
But…must enter data correctly according to your data dictionary
Critiquing stats
- Identify if descriptive, parametric, or nonparametric
- Identify the level of measurement for the I.V. and D.V. of each hypothesis
- check the chart in book to see if the statistical test was appropriate for each hypothesis
- check alpha and p value
- check significance and non-significance of results
- check CI if provided
- Check for good charts or tables