continuous outcome

Question 1

continuous outcome, non adjusted tests and adusted tests?

Answer 1

non adjusted
> t test
> Wilcoxon Mann-Whitney test

adjusted
> linear regression
> ancova

Question 2

what is used to determine treatment effect with a continous outcome

Answer 2

mean difference

Question 3

prior assumptions of a two sample t test?

Answer 3

• Groups are independent
• Individual measurements are independent
• Outcome is normally distributed

Question 4

what are covariate effects

Answer 4

the quantified baseline covariates

Question 5

what do we call an analysis based on covariate effects

Answer 5

ANCOVA

analysis of covariance

Question 6

prior assumptions of linear regression

Answer 6

• Errors (residuals) are independent
• Errors are normally distributed for each value of X
• Errors have the same variance for each value of X (homoscedasticity)
• residual plot (plotting the residuals against a continuous predictor)

Question 7

How do we check the assumptions of normality in the data?

Answer 7

• Histogram of residuals (should see gaussian distribution)
• Qq plot of residual (residuals should fall on a straight line)
• plotting the residuals against the fitted value (should see random scatter around zero)

Question 8

what is a Q-Q plot of residuals?

Answer 8

the standardized residuals are plotted against the expected normal quantiles (z-scores)

Question 9

why would we plot the residuals against the fitted value?

Answer 9

To identify potential issues with the model, such as nonlinearity, heteroscedasticity, or outliers.

If the points on the plot are randomly scattered around zero, = suggests model is a good fit for the data.

If the points show a clear pattern e.g., U-shape, suggests that the relationship between the predictor and outcome variable is nonlinear, or that there is a problem with heteroscedasticity (i.e., the variability of the residuals is not constant across the range of the predictor variable).

Question 10

what is a residual plot used for?

Answer 10

helps identify potential nonlinear relationships between the predictor and outcome variables.

plot residuals against continuous predictor

If relationship is linear, the residuals should be randomly scattered around zero, with no clear pattern. However, if the relationship is nonlinear, the residuals will show a pattern that suggests a deviation from linearity.

Question 11

why is it a good idea to check the histogram of continuous variables before reporting them

Answer 11

to determine which summary statistic is best for reporting central tendancy and variability

• variable with normal distribution, report mean and SD
• skewed distribution, report median and interquartile range (IQR)