Inferential Statistics Flashcards
What are the characteristics of research study?
- Formulate. Research question/study design/generate hypothesis
- Patient recruitment, data collection and entry
- Data analysis
- Descriptive statistics
- Inferential statistics
What are the 2 types of measurement?
Numerical and categorical
What must you decide before undertaking before undertaking any analysis?
- are the data evenly distributed?
- is it parametric or non-parametric?
What are the 4 conditions that must be fulfilled to undertake parametric test?
- data must be interval/ratio
- subjects must be randomly selected
- data must be normally distributed
- variation in results from each condition should be similar
What is descriptive data analysis?
- Subjects usually only measured once
- establishes only associations between variables
What is experimental data analysis?
- subjects measured before and after a treatment
- can establish causality
How do you establish a p-value?
- done before the start of the study
- use a power calculation to determine sample size needed
What are the 2 types of study design?
Experimental and correlational
What is Chi-square test?
- considers the distribution of actual data and what would be expected
- DV is categorical
- based on frequency counts
- minimum response size = 80% of cells must contain a minimum expected frequency of 5 and no empty cells
What is considered a weak r value?
Between 0.1-0.4
What is considered a moderate r value?
0.5-0.7
What is considered a strong r value?
0.8-1.0
What is a correlational coefficient?
- a numerical description of this tendency
What does a -1 correlation coefficient mean?
Perfect negative correlation
What does a +1 correlation coefficient mean?
Perfect positive correlation
What does a correlation coefficient of 0 mean?
No relationship
What are he conditions for correlations?
Need paired sets of measurements from either:
- 2 different sets of measurements from the same sample
- same set of measurements from matched pairs of individuals
- same set of measurements from same individuals at 2 different points in time
A linear relationship
Which statistical test do you use for a correlational parametric data?
Pearson product moment
What are the characteristics of Pearson product moment?
- all data is continuous
- data is normal
- relationship is linear
What test do you use for a correlational non parametric data?
Spearman rank correlation
What are the characteristics of Spearman rank correlation?
- both variables from the same sample or matched pairs
- either ordinal data or skewed continuous data
How can you tell is a result is significant?
- p>0.05 = not statistically significant
What is the confidence interval (CI)?
- the range of values within which it is reasonably certain that the population value lies
- concerned with precision, how much the sample estimates is likely to differ on average from the true population value
- can be calculated for different tests
- estimate the effect of treatment
- width of the CI gives an indication of the precision of the estimate
- if statistically significant the CI won’t include the null hypothesis
Which test do you use for 2 groups and parametric tests?
Unrelated t test
What tests are used for between group comparisons?
- unrelated t test
- Mann-Whitney U test
- fishers exact test
- one-way ANOVA
- kruskall Wallis
- Chi-squared test
What tests do you use for within group comparisons (repeated measures)?
- related t test
- wilcoxon signed rank test
- McNemar
What statistical test do you use for 2 groups and non-parametric test?
Mann-Whitney U test
What statistical test do you use for 2 groups and categorical data?
Fishers exact test
What statistical test do you use for >2 groups and parametric tests?
One way ANOVA
What statistical test do you use for >2 groups and non-parametric tests?
Kruskall Wallis
What statistical test do you use for >2 groups and categorical data?
Chi-square test
What statistical test do you use for 2 measures and parametric tests?
Related t test
What statistical test do you use for 2 measures and non-parametric test?
Wilcoxon signed rank test
What statistical test do you use for 2 measures and categorical data?
McNemar
What are the characteristics of unrelated t test?
- compares the means between groups
- looks at descriptive data: normal distribution, continuous scale, unpaired data, standard deviations should be similar
What are the characteristics of Mann-Whitney U test?
- non-parametric equivalent of t test
- compares ranked scores for each group
- independent groups
- ordinal or interval data
- small or unequal samples
What are the characteristics of one way ANOVA?
- single dependent variable
- continuous/interval scale
- normal data with similar standard deviation
What are the characteristics of Kruskal-Wallis test?
- interval or ordinal
- group differences in the mean rank values
What are the characteristics of related t test?
- within group comparisons
- repeated measures with same individual
- measurements from matched (demographically) pairs
What are the characteristics of Wilcoxon signed rank test?
- data for each group related matched pairs or repeated measures
- continuous or ordinal scale with at least 4 points
When testing an idea or hypothesis what do you need to do?
- a hypothesis must be devised and stated
- a researcher project must be designed which tests the hypothesis
- results must be analysed with descriptive and statistical test
