Bell's Palsy Clinical Trial

Bell's Palsy is a condition that causes temporary paralysis on one side of the face. A clinical trial explored different treatments for the condition and tracked patient factors like age, time since onset, and response severity.

This homework exercise uses the dataset from that trial. Please load the dataset into Gretl and create cross-tabulations of:

• recovery vs. non-recovery at 3 months by treatment
• recovery vs. non-recovery at 9 months by treatment

Using those cross-tabulations, answer the following questions:

• which treatment is most effective at 3 months? at 9 months?
• which treatment is least effective at 3 months? at 9 months?

Let's assume that patients prefer to recover in 3 months. Create cross-tabulations of:

• recovery vs. non-recovery by gender
• recovery vs. non-recovery by age

Using those cross-tabulations, answer the following questions:

• how does gender affect recovery?
• how does age affect recovery?

Advanced Topic: Probability Models

If you continue your study of statistics, you'll learn about probability models, which allow us to use multiple variables to predict the probability of recovery. With such models, we can estimate the effect of one variable, while holding the others constant.

Probability models are outside the scope of an introductory course. Nonetheless, Gretl makes it easy to estimate them.

So go to the Model menu, select Limited dependent variable, then select Logit and finally select Binary. A new window will open where you can specify the model to estimate.

In the new window, set full_3month as the dependent variable. As regressors, select female, age, hbmod_base, days_until_treat, prednisolone and acyclovir. And in the options at bottom, select Show slopes at mean.

The regressors and estimated coefficients form a linear model that generates a score. A higher score corresponds to a higher probability of recovery.

Because the probability is a number between 0 and 1, the function's slope changes as the values of the variables change. So to evaluate the effect of a given variable, we want to examine the function's slope at the sample means of the variables. These are the values in the rightmost column (labeled: "slope").

Using the slopes, answer the following questions:

• by how much does treatment with Prednisolone affect the probability of recovery?
• by how much does treatment with Acyclovir affect the probability of recovery?