Week 10 - Chapter 23 Statistical interference and statistical conclusion validity Flashcards
What is a type 1 error:
Mistakenly finding a difference
- rejecting the null hypothesis when the null hypothesis is true
What is a type 2 error
Mistakenly not finding a difference when there is a significant difference
- do not reject null hypothesis when null hypothesis is false
Level of significance
represents criterion for judging if an observed difference can be considered sampling error
- typically 5%
- at the 5% level of significance there is a 5% of rejecting the null hypothesis when it is true
What is power
- can be though oof as sensitivity
- the more sensitive a test the more likely it will detect important clinical differences that truly exist
- used to avoid type 2 error
Effect size index
- unites standardized value that allows for comparisons across samples and studies
- power is influenced by the size of the effect of the independent variable
Z-ratio/Z-score
represents the distance between that score and the sample mean divided by the standard deviation of the distribution
Critical region
- this is the region of rejection for H0
- the area above dn below a z-score
Two-tailed test
- critical region is split by the noncritical region
- above and below
critical value
- the value of Z that defines the critical region
one tailed test
situations where a researcher has sufficient reason to porpoise an alternative hypothesis that specifics which mean will be larger a directional test can be preformed
Statistical power
the power of a statistical test concerns its ability to document a real relationship between independent and dependent variables
Violated assumptions of statistical test
most statistical procedures are based on a variety of assumptions about levels of measurement and variance in the data or sample from which they are collected
Reliability and variance
statistical conclusions are threatened by an extraneous factors that increase variability within the data
- such as unreliable measurement, failure to standardize the protocol, environmental interferences or heterogeneity of subjects
failure to use intention to treat analysis
- when data are missing or subjects do not adhere to random group assignments, analysis procedures are designed to maintain the important element of randomization
Probability
- likelihood that any one event will occur (ratio and decimal)
- signified with Lowe case P
what are the issues of probability research
- implies uncertainty: predicts what should happen not what will happen
- how well does sample estimate characteristics of population: large sample size = greater probability of sample estimating population
- determine if observed effects likely to have occurred bye chance
Confidence intervals
- use of probability to estimate population mean in 2 ways, point estimate and confidence interval
point estimate
uses sample mean to estimate population mean
- can include error
How to minimize error in confidence intervals
- calculate ranges of scores that should contain the population mean (confidence interval)
- boundaries of range calculated form sample mean, standard deviates and sample size
- expressed as probability percentage
Normal sample distribution in a confidence interval
- Z score = +/- 2
- 95% of population within 2 standard error units of mean
(hypothesis testing
- decide if observed effect is likely due to chance variation (hypothesis testing)
Null hypothesis
- observed difference between groups occurred by chance
Alternative hypothesis
- observed effect too large to be result of chance alone
Non directional alternative hypothesis
mean change between treatment A and treatment B will be different
Directional alternative hypothesis
change for treatment A will be larger than change for treatment B
Factors affecting Power
- P= power: can be calculated after statistical analysis or during planning
- A = level significance: as a decreases then power decreases
- N= number of subjects; as sample size increases power increase
- E= effect size; as effect size increases power increases (how big of an affect does the intervention have)
Z-ratio
- normal distribution can determine area beyond any point
- calculated Z-score for sample
- critical value: points on scale beyond which on 5% of values would be
- Z-score: for sample not in critical region there is no significant difference
Statistical conclusions: how do we come to conclusions
1, p>alpha
2. p<alpha
- p-value is the probability occurred by chance
- probability (p) > a, not statistically significant and do not reject H0
- probability (p)<a there is a statistical significance difference and you accept HA
Parametric statistics
- used to estimate population parameters
- validity of parametric tests depend on parametric assumptions:
1. samples randomly drawn from population with normal distribution
2. variances o samples being compared are roughly equal
3. data measured on interval or ratio scales - if parametric assumptions are unable to be met uses nonparametric