# Thread: Two therapies in two separate populations over time

1. ## Two therapies in two separate populations over time

Good Evening,

I would appreciate any input on the following problem, I have searched around the web to see if I can find a more effective statistical measure to use to evaluate these two populations to no avail. Even confirmation that ANOVA/sample mean analysis is the only way to look at this would be appreciated!

There are two patient populations, difference between them being that they are being treated with two different therapies. Their blood work samples are sent to two different labs for evaluation (each population is evaluated by both labs). We have about 80 patients - with data points available semi-annually.

What we wish to assess is whether the two labs that provide the results are statistically different. If so, how different? I would like to control for the different therapies that each patient is on while assessing for mean differences.

Course of Action (at this point):
Code therapy 1; therapy 2
Enter all data available over time with results coded as lab 1 lab 2
Conduct ANOVA (Univariate) with independent variable as Lab, dependent as Results - add in a control for therapy.
Problems: Normal distribution of dependent variable. Controlling for outliers. Unsure whether most efficient method.

Since the results can vary significantly between patients as well, should I control by patient number as well? As for the outlier problem, I can account for that by conducting a data cleaning for data points used but was wondering if there is a better test to evaluate this. Any help would be very much appreciated! =)

Bv

2. ## Re: Two therapies in two separate populations over time

What sample size are talking about? Seems like with the number of variables your interested in, perhaps general linear regression or regression controlling for clusters. How many multiple data points are in the dataset.

3. ## Re: Two therapies in two separate populations over time

Thank you for the reply! =)

Sample sizes are currently an estimate:

Patients (approx): 80
Data points per year: 2
Average # of years: 8
80*2*8 = 1280 data points to assess.

One of the bigger issues regarding this analysis is that the blood test that we are looking at has a huge range and inter-patient variability. Hence, I was not sure if regression would work or not =( I am probably wrong in this assumption, please feel free to point this out! =)

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts