Sample size estimation with missing values on clinical trials
Missing data is a common problem that with extremely damaging inferences from clinical trials. This unavoidable defect is mainly due to human factors, the patient being unable to follow up and some types of observations going missing (Piantadosi, 2005). An example would be the patient refuses to und...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2019
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/136483 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Missing data is a common problem that with extremely damaging inferences from clinical trials. This unavoidable defect is mainly due to human factors, the patient being unable to follow up and some types of observations going missing (Piantadosi, 2005). An example would be the patient refuses to undergo an examination. In this work, we compare the power of the full datasets with the power of the datasets having missing values under the 2 × 2 and 2 × 4 crossover design. In this way, we find the average percentage of the sample size that should be increased to offset the decreasing in power (original is around 80%), which is caused by 20% and 30% missing data in the tests for treatment difference and bio-equivalence. Also comparing the number of subjects (full data) which can reach 80% power through formulas and simulations. The aim of this comparison is to distinguish whether the general practices would have an overestimate or underestimate for the chosen mixed model in the analysis. |
|---|