SPMM - Statistics

Question 1

What is the coefficient of variation useful for?

Answer 1

Comparing variations of two or more different variables

Question 2

How is degrees of freedom calculated in chi square statistics?

Answer 2

df = (number of rows - 1) x (number of columns - 1)

Question 3

What does chi square test?

Answer 3

Observed vs expected proportions

Question 4

What does cox’s proportional hazards test asses?

Answer 4

If survival time (continuous) is affected by one or several predictors

Question 5

What type of data does ANOVA use?

Answer 5

ANOVA assess for differences in a continuous dependent variable (interval or ratio) based upon several categorical (nominal or ordinal)

Question 6

When would McNemar test be used?

Answer 6

Dependent is dichotomous
Independent is categorical

Similar to paired t test but for dichotomous rather than continuous dependent variable
Like Chi Square but for paired + non-parametric

Question 7

In a logistic regression how may a chi square be useful?

Answer 7

Allows to test the null hypothesis that all covariates in a logistic regression are 0

A significant result suggests that one of the covariates is significantly associated with the dependent variable

Question 8

When is Wilcoxon test used?

Answer 8

For ordinal paired data that is non-normally distributed
- Two types exist wilcoxon rank sum for larger samples
- Wilcoxon sign for smaller ones

Both calculated by converting data into ranks and summing the positive and negative ranks in a table

Question 9

When is a log rank test used?

Answer 9

To see if survival differs significantly between two groups

Assess significant differences from the null hypothesis which is the probability of death between groups is equal

Question 10

How do independent groups t test and Mann Whitney U tests differ?

Answer 10

Independent groups/unpaired t test - if the data is normally distributed and continuous

Mann Whitney U - non-parametric for not normally distributed data OR ordinal data as this would utilise medians rather than testing of means

Question 11

Name the assumptions of chi square?

Answer 11

All expected values in each cell have a count = 1 (no cell expected value of < 1)
At least 80% of total cells have a count = 5

Independent groups
Data must be counts or frequencies rather than proportions or percentages
All subjects can only contribute to one cell/data point
Categorical data (can be ordinal - interval/ratio would need to be converted)

Question 12

When may weighted kappa be chosen > Cohen’s kappa for inter-rater reliability?

Answer 12

Ordinal data - it takes into account degree of disagreement and frequencies of disagreement

Question 13

What method corrects/adjusts for multiple comparisons?

Answer 13

Bonferroni correction - it is rigorous and strict and may lead to false negatives

Question 14

What is censored data in kaplan meier survival curves?

Answer 14

Events that occur after the period witnessed.

In Kaplan meirer survival curves may only plot until 50% of the sample reach outcome of interest (median survival time)

For the remaining 50% of participants who don’t reach the outcome in the study period but may reach it after this data is referred to as censored data

Question 15

Describe when a one-way ANOVA would be used?

Answer 15

One categorical independent variable (N-way ANOVA is where there is N number of categorical independent variable)

Dependent variable is continuous

Question 16

Spearman’s rank correlation may be chosen over pearson’s if?

Answer 16

Discrete data
Continuous but non-parametric

Question 17

How does stepwise, backward and forward regression analysis differ?

Answer 17

Stepwise:
- Start with ones which are least statistically significantly associated with the outcome - from here more statistically significant are added
- Disadvantage is that clinically relevant variables may not be statistically significant

Forward:
- Construct the regression with just independent and dependent variables
- Add each covariate - if the regression coefficient of a previously added covariate changes then the covariate must be a confounder and is retained (if not it is discarded)

Backward:
- Starts with the full model and removes each covariate depending on changes seen in the regression coefficients

Question 18

How is kappa calculated?

Answer 18

Observed agreement beyond chance / potential agreement beyond chance (takes into account some agreements may be chance)

Question 19

How is observed agreement calculated?

Answer 19

No. observed agreement of presence + No. of observed agreement of absence

/

Total number of presence + total number of absence

Question 20

What does R2 in regression models mean?

Answer 20

Coefficient of determination - the amount of variation in a model that is account for by the regression model