Biostatistics | Midterm - HYPOTHESIS TESTING Flashcards
A statement about one or more populations.
- It is usually concerned with the parameters of the population. e.g. the hospital administrator may want to test the hypothesis that the average length of stay of patients admitted to the hospital is 5 days
HYPOTHESIS TESTING
a statement of the research question that sets forth the appropriate statistical evaluation
Hypothesis
statement of no differences or association between variables
Null hypothesis “H0”
statement of differences or association between variables
Alternative hypothesis “H1”
STEPS IN HYPOTHESIS TESTING
- Make a prediction
- Decide on a statistical test to use
- Select a significance level and a critical region (region of rejection of the null hypothesis). To do this you must consider two things.
- Whether both ends (tails) of the distribution should be included.
- How the critical region of a certain size will contribute to Type I or Type II errors.
- Computing the test statistic - The test statistic is not a mean, sd or any form of descriptive data. It is simply a number that can be compared with a set of results predicted by the sampling distribution.
- Compare the test statistic to the sampling distribution (table) and make a decision about the null hypothesis reject it if the statistic falls in the region of rejection
- Consider the power of the test – its probability of detecting a significant difference – parametric tests are more powerful
Testing a Hypothesis:
outcome is expected in a single direction (e.g., administration of experimental drug will result in a decrease in systolic BP)
One-tailed hypothesis
Testing a Hypothesis:
the direction of the effect is unknown (e.g., experimental therapy will result in a different response rate than that of current standard of care)
Two-tailed hypothesis
Statistical Errors:
False Positive
Type 1
Statistical Errors
False Negative
Type 2
Under the assumption that H0 is correct, probability that the observed test statistic is skewed towards H1
Probability Value (p-value
Confidence Intervals:
90%
p<1.0 (0.5 for two-tailed)
Confidence Intervals:
95%
p<0.05
Values:
the set cut-off value from statistical table (e.g. t-test 1.96 = 95% CI; z-test 2.58 for two tailed
Critical
Confidence Intervals:
99%
p<0.01
Test to use for Analysis of Continuous Data
Correlation t-test, T-test
Values:
the value from the data collected
Computed
Test to use for Analysis of Categorical Data
Chi-square, Logistic regression
Types of DATA ANALYSIS
Analysis of Categorical Data
Analysis of Continuous Data
Measures of Association
PARAMETRIC & NON-PARAMETRIC
PARAMETRIC
Normality
Skewness/ kurtosis
Linearity
Heteroscedasticity/ homoscedasticity
Correlation / Test of Relationship
- Parametric Test
Pearson r
Correlation / Test of Relationship
- Non-parametric
spearman-rho
Correlation / Test of Relationship
1.Tests
a. Parametric - pearson-r
b. Non-parametric – spearman-rho
c.Chi-square
Note: if data is encoded as frequency – indicate data weights
d. Kendall-Q for nominal data
e. Phi-coefficient – ordinal data
Measure of relationship
a. Direct
b. Indirect
Strength of relationship
a. -1 to +1; 0 = no correlation
b. Scatter plot
Interpreting strength of correlations:
0 – 0.25
Little or no relationship
Interpreting strength of correlations:
0.25 – 0.50
Fair degree of relationship
Interpreting strength of correlations:
0.50 - 0.75
Moderate degree of relationship
Interpreting strength of correlations:
0.75 – 1.0
Strong relationship
Interpreting strength of correlations:
1.0
perfect correlation
- concerned with the effect of one variable to the relationship of 2 variables
- Relationship of two variables while holding the third variable constant. (eg. Science and Reading while holding IQ constant)
Partial Correlation
Test of Difference:
>30 samples; large sample size
T-test = <30 samples (homogenous) – tested by Levene’s test of homogeneity; hypothesis = sample is not homogenous
Z-test
Test of Difference:
<30 samples (homogenous) – tested by Levene’s test of homogeneity; hypothesis = sample is not homogenous
T-test
