PAS - Data Analysis

Question 1

what is a t-test?

Answer 1

statistical test that compares means to determine if there’s a statistical difference between two groups

data is assumed to be parametric - follows a normal distribution

Question 2

statistically significant p value?

Answer 2

when the p value is smaller than the alpha value/ threshold

reject null hypothesis (Ho)

Question 3

when the p value is greater than the alpha value/ threshold (0.05)…

Answer 3

no stastically significant difference

accept null hypothesis (Ho)

Question 4

risk of smaller alpha value threshold?

Answer 4

higher risk for false negatives/ type 2 errors

Question 5

risk of larger alpha value?

Answer 5

higher risk of false positives/ type 1 errors

Question 6

paired t test?

Answer 6

comparing measurements taken from the same subjects at different timepoints

e.g. blood pressure measurements before and after treatment from the same person

Question 7

unpaired t test?

Answer 7

comparing measurements taken from two independent samples/ subjects

e.g. blood pressure measurements forma control vs treatment group

Question 8

parametric data?

Answer 8

data assumed to follow a normal distribution

Question 9

non-parametric data?

Answer 9

skewed data - doesn’t follow a normal distribution

small sample sizes are often non-parametric data

Question 10

how to identify if data is parametric?

Answer 10

1. evaluate data distribution by plotting the data - e.g. histogram plots
2. test of normality - e.g. Shapiro-Wilk test - p < 0.05 = sufficient evidence to suggest non-parametric data
3. assess sample size - small sample sizes will likely produce non-parametric data

Question 11

statistical test conducted with parametric & paired data?

Answer 11

paired t-test

Question 12

statistical test conducted with parametric & unpaired data?

Answer 12

unpaired t-test

Question 13

statistical test conducted with non-parametric & paired data?

Answer 13

Wilocon signed rank

Question 14

statistical test conducted with non-parametric & unpaired data?

Answer 14

Mann-Whitney U test

Question 15

one-tailed test?

Answer 15

test assessing the possibility of effect in one direction between two groups

e.g. treatment causing only an increase in survival

Question 16

two-tailed test?

Answer 16

test assessing the possibility of effect in both direction between two groups

e.g. treatment causing an increase or decrease (any change) in blood pressure between two groups

Question 17

risk of conducting multiple statistical tests on data?

Answer 17

increases chance of false positives/ type 1 errors

Question 18

what is ANOVA?

Answer 18

analysis of variance - statistical analysis test to analyse the differences between means of more than two groups, make multiple comparisons

assume data is parametric & independent/ unpaired samples

Question 19

one way ANOVA?

Answer 19

compares the difference between means of three or more independent groups on a single continuous variable

e.g. three independent groups & the effects of treatment on blood pressure

Question 20

two way ANOVA?

Answer 20

compares how the mean of two quantitative/categorical variables change with a single continuous variable

e.g. sex of subjects within sample (two categorical variables) & effects of treatment on their blood pressure

Question 21

statistical type of ANOVA test for unpaired, parametric data?

Answer 21

one way ANOVA

Question 22

statistical type of ANOVA test for paired, parametric data?

Answer 22

repeated measures ANOVA

(tests whether there are statistically significant differences in three or more dependent samples)

Question 23

statistical type of ANOVA test for unpaired, non-parametric data?

Answer 23

Kruskal-Wallis test

Question 24

statistical type of ANOVA test for paired, non-parametric data?

Answer 24

Friedman test

Question 25

what is a categorical variable?

Answer 25

variables with discrete options/states with no predictable quantitative scale/ relationship between them

e.g. success/failure, hair colour, race, sex

Question 26

what is a continuous variable?

Answer 26

variables measured on a quantitative scale

e.g. age, height, weight, temperature, blood pressure measurements

Question 27

what are post-tests?

Answer 27

evaluates specific pairwise comparisons following significant ANOVA tests results

Question 28

what are the three types of post-tests?

Answer 28

1. Tukey’s test
2. Punnett’s test
3. Bonferroni correction

Question 29

Tukey’s test?

Answer 29

pairwise comparison post-test following ANOVA for EVERY possible combination of groups

e.g. A compared with B, C, D; B compared with C, D,; C compared with D

(all combinations of all groups = p-values for each combination are calculated)

Question 30

Punnett’s test?

Answer 30

pairwise comparison post-test following ANOVA for each group to a control group

e.g. A(control) compared with B, C, D
= p-values for each combination are calculated and compared

Question 31

Bonferroni’s test?

Answer 31

pairwise comparison post-test following ANOVA for any chosen combination (not adhering to a pattern)

e.g. p-values calculated and compared between B and D

Question 32

what are the three statistical tests for analysing categorical data?

Answer 32

Pearson’s Chi-squared test
Fisher’s exact test
McNemar’s test

Question 33

Pearson’s Chi-squared test?

Answer 33

for large sample sizes and 2x2 or larger contingency tables with unpaired/ independent data

tests the null hypothesis - no significant difference between the expected and observed frequencies

Question 34

Fisher’s exact test?

Answer 34

for small sample sizes (or when expected frequency for a category are less than 5)

ideal for 2x2 contingency tables with unpaired/ independent data

Question 35

McNemar’s test?

Answer 35

for categorical data in 2x2 contingency tables with paired data

determines if there’s a significant difference in proportions for matched pairs

Question 36

types of data for Chi-squared tests analysis?

Answer 36

categorical data
independent samples (unpaired)
not a 2x2 contingency table or small sample size - analyses data that doesn’t meet Fisher’s exact test requirements

Question 37

types of data for Fisher’s exact test analysis?

Answer 37

categorical data
independent samples (unpaired)
2x2 contingency table or small sample size (under 10 samples)

Question 38

types of data for McNemar test anaylsis?

Answer 38

categorical data
paired data (non-independent)
conforms to 2x2 contingency table

Question 39

ideal post-hoc tests for multiple comparisons of categorical data (i.e. with 2x3 or more contingency tables)?

Answer 39

chi-squared or Fisher’s exact tests with Bonferroni correction

adjusts p values to account for multiple comparisons by multiplying the raw p-value by the number of comparisons (p x n)

helps control type 1 errors/ false positives

Question 40

contingency table?

Answer 40

displays frequencies/outcomes for two categorical variables

Question 41

limitations of p-values?

Answer 41

p-values assess whether observed differences are likely due to random variation or actual differences between groups - don’t measure the practical significance or magnitude of the difference

e.g. a statistically significant p-value result might only represent a minor difference that isn’t practically meaningful

Question 42

confidence interval?

Answer 42

alternative to p-values

gives a range of values within which the true effect size is expected to lie with a certain probability

e.g. a 95% CL means the true mean difference between groups would lie within that mean range 95% of the time

Question 43

compare p-values and confidence intervals?

Answer 43

p-value = indicates the probability of observing the data if the null hypothesis is true
- p value smaller than the alpha threshold suggests there’s a greater chance of the observed data being statistically different to the null hypothesis/ expected data

confidence interval = provides a range in which the true (mean) difference between two groups is expected to lie within the given CL % value

Question 44

statistical significance, p-values confidence intervals? - based on 0

Answer 44

if a 95% CL doesn’t contain 0 within its range - the result is statistically significant at α = 0.05
- implies p < 0.05

if a 95% CI contains zero - the result is NOT statistically significant at α = 0.05
- implies p > 0.05

if 95% CL has 0 as a boundary - p = 0.05