Chapter 14 Flashcards
5 levels of measurement
Nominol, ordinol, interval, Ratio, Continous varaible
Nominol measure
lowest level that involves number to desginate attributes
- numeric value is not mathmatic
Ordinal
Rank people based on ADLs
1- dependent 4 independent
interval
Rank people on attributes
- with a number that clearly specifies a distance btwn the 2
IQ is an example
Ratio
- are the highest level
- numbers can hava meaningful zero
ex: weight
Continuous variable
- have both internal and ratio measurements
Descriptive Statistics
- used to synthesize data
Parameter
calculated values averages and percentages
Statistic
descriptive index from sample
Freq distributon
- set of values form lowest and highest includes a count
Symmetric Distribution
- if a grpah is folded in hald it would superimposed
Skewed Distribution
- Majority of data peaks to one side
Normal
bell shaped curve lower peak
Central tendency
- include methods to determine a central value
Mode- most frequent
MEdian - divides score in half
MEan - the average
Varaibilty
Central tendecy is the same for two different sets of distributions
RAnge
High subtract from low
Standard devaition
variability index calculated by every value in a distribution
-Averag amount
In a normal bell shaped curve
1 SD= 68%
2 SD = 95%
3 SD = 99.7 %
Bivariate descriptive statistics
if frequency dsitribution is based on 2 variables
Correlation
RElationship betwn 2 variables
- coefficient is intenist and direction (-1 and +1)
+ is positive
- pearosn’s r, spearman row, correlation matrix
Absoluete risk
measures the undsirable outcome
Absolute risk reduciton
consider differences between 2 groups
Odds ratio
results of undesirable effect between 2 groups
Number needed to treat
estimates how many people need to be treated fro one desirable outcome
1 / absolute reduction
Inferential Statistics
uses law of probability to test research hypotheses
- Larger sample size helps reduce chance of outlying data
- SEM: standard deviation of mean of smapel higher the number more error
-
Parameter estimation
used to estimate a pop parameter such as mean
- Confidence interval: relates to probability of being right 95% or higher indicates corectness
Hypothesis tetsing
- uses objective criteria to see if hypothesis is supported
TYPE 1 error
reject null hypothesis that is true
Type 2 error
when null hypothesis is rejected but the independent varaible had an outcome on the dependent variable
Level of Signficance
-indicated by alpha
- is confidence interval subtracted by 1
- 1- .95 =0.05
Power analysis: calculated to determin chancee of type 2 error power of at least .80
Stattistical signficane: results are not due to chance fall within 2 deviations of mean
Nonsignifacant: results could be due to chance
Specific Statistical Test T-test
T-test: Parametric test for testing differeces in 2 groups individual value is not importatn
P-Value: Determines if results are significant
P-value of less than 0.05 indicates statistical significance
Independent T test: appropriate fro 2 groups
PAired: one group is tested at 2 different times
ANOVA
calculates an F-value
- varies for each study
- show that interventions are effective for a p-value less than 0.05
Chi- Square test
focuses on different in proportion
-gives a p-value
Multiple REgression
meaures several independent variables
-Predictors
- R (0-1), higher R square value more likly predictor accounted fro variation
higher R squared
ANCOVA
Effective when control is not acheived through randomization
- Combo of ANOVA with multiple regression
- control covariance of cofounding variables
Logistice Regression
Analyzes relationship between multiple independent vraiblaes
- yeild Odds ratio and confidence interval
intraclass correlation coefficient
- used for test- retest reliability
- score closer to 1 means stronger reliability
Cohen’s kappa
interrater reliability
- determines if two individuals will write something similariy
Coefficient alpha or Chronbach’s alpha
is used to measure how often components of a multicompnent tool meaure the same attribute
Conetn validity
determines if content relates to construct of interest
0.9 or higher is good
Criterion validity
- concerns extent to which scores on a measure are consistent with gold standard
- sensitivity is ability to diagnose a condition
- specificity measure ability to screen out those with a condition
Construct validity
concerns extent to which a measure was truly measured in trage tconstruct
- pearsons r value or independt t -test
