We discussed some of the limitations of the "Acupunture for Migrain" paper in class last week. Please provide a short report that would be aimed at the authors of the paper that would be helpful in them designing another such study or in the analysis and reporting of the study reported in the articular. Target length: 450 to 550 words.
Statistical Analyses and Methods in the Published Literature: The SAMPL Guidelines
Papers relating to presenting results of complex model estimation
Q1. The files dose.csv and lab.csv contain records of patients being treated to prevent blood clots by infusing an anti-coagulant into a vein. The rate of infusion may be changed and infusion may be temporarily halted. The dose file contains one line for ever time the infusion rate was changed, recording the new infusion rate and the date and time of change. Temporary halts are indicated with an infusion rate of 0. Termination of infusion is indicated by 0d. The lab file contains a lab measurement made on a blood sample. Lower values indicate that the blood is more prone to clotting.
The clinical investigators wished to examine the temporal relationship between anti-coagulant dose and lab test values. Starting Thursday and continuing to Tuesday's class, we will discuss how this might be done. Your task for the current assignment is to "munge" the data to support such an analysis.
Q2. A clinical trial was conducted in which patients were randomized to receive anti-coagulant therapy for 5 days versus 10 days to prevent clots.
A lab test as in the above example was done at baseline as a potentially important covariate. Suppose that the log odds risk for developing a clot is 5 - LabTest/8$ for the short group and 2 - Labtest/8 for the long group. Conduct simulations of a study with 250 patients in each arm to compare odds ratio estimates for the treatment effect with or without covariate adjustment for baseline lab value. In your simulation use the first recorded lab value for the subjects in the lab file to represent the distribution of baseline values. Comment on the difference in the odds ratio estimates. Should the lab value be regarded as a confounder?
"Discover" 3 types of bias and provide short definitions, posting to the Bias (+ Confounding?) thread on the course WIKI by noon Thursday.
Quality-of-Life Assessment: Can We Keep It Simple? by D. R. Cox, R. Fitzpatrick, A. E. Fletcher, S. M. Gore, D. J. Spiegelhalter and D. R. Jones provides very good discussion relating to development of "scales" based on questionnaires. I strongly suggest at least bookmarking this for later reference.
Please review the project description and post three questions for discussion on the wiki page by noon, March 4.
Related readings
Conduct a descriptive analysis of the data, using graphs and tables, to provide informal answers to the investigators questions (i.e. essential summary statistics, but no p-values, confidence intervals.)
Conduct a preliminary analysis of the using the change from the pre-dialysis to the first dialysis BP readings following the Hill's and Armitage approach described in this excerpt from Practical Statistics for Medical Research, by D.G.Altman
Rather than using change scores, it would be possible to follow an analysis of covariance approach. Conduct some simple analyses to provide guidance as to the potential gain in efficiency from such an analysis.
Please consider the three examples described in this handout and post three questions by noon this Thursday for discussion that day.
I've extracted a subset of the questions posed on the WIKI regarding the Dialysis study. Please review these questions and come to class prepared to contribute answers to at least two of the questions. We'll also cover the other questions raised. As well, you may wish to add additional questions to the WIKI in relation to Assignment 2 (see below).
Case study: Please review the project description and sample data extract and post three questions on the for discussion on the wiki page by Wednesday evening.
Description of Split Plot Exeriment
Analysis of Split Plot Exeriment
Study designs to be presented
Re-read BMJ paper noted below in preparation for class discussion about incorporation of baseline observation in Asthma study.
Planning the Statistical Analysis: Here is a summary of the questions that were posed last week in relationship to sample size considerations for the Asthma study. Choose one of the questions under //E: Analysis Strategy, Covariates// and come to class Thursday prepared to discuss how the answer to that question will affect the statistical analysis plan. Some of the questions in the other sections (i.e. parts A, B, C, D etc) are also relevant to determining the anlaysis plan. Pick one of those and be prepared to to discuss it as well. Please post the question numbers you've selected on the WIKI (e.g. E3, D2).
Read:
For Discussion:
Activated carbon is used to treat poisonings and overdoses following oral ingestion as it is thought to bind the poison and prevent its absorption by the gastrointestinal tract. A group of emergency physicians wished to assets it's benefits in comparison with other treatments comparing effectiveness, defining 24 discharge from hospital at or before 24 hours from admission as a positive outcome. There plan was to conduct a retrospective review of charts of children admitted with poisoning to compare the relationship between use and non-use of charcoal and positive outcome rates. Since manual chart reviews are time consuming and expensive, they wished to determine an appropriate sample size for making this comparison.
A relevant sample size calculation may be made using a standard two-sample approach for comparing proportions - for example see: Comparing Proportions for Two Independent Samples
On Thursday's class we will discuss the information requirements for doing this calculation. To review sample size calculations in general see this chapter from Gerald van Belle's book below: Please visit the STAT 550 wiki discussion page to add to the discussion.
Conduct appropriate analyses and write brief reports short report, emphasizing descriptive statistics, for two (your choice!) examples in this handout
Asthma is a disease which causes inflammation in an individual's respiratory airways (larynx, trachea, bronchi, and bronchioles), leading to symptoms such as wheezing, coughing, chest tightness, and shortness of breath. It is a chronic disease (there is currently no cure for asthma) but it's effects can be controlled by regular inhalation of drugs known as bronchodilators, which reduce the inflammation. Treatments are self-administered regularly each day.
One aspect of assessing the effectiveness of treatment is to by measurement of lung function use an instrument known as a spirometer. One commonly used measurment is Forced Expiratory Volume in 1 second (FEV1), which is the volume of air that a patient can exhale when asked to exhale as forcefully as possible. FEV1 is reduced in asthma sufferers but treatment with bronchodilators can restore FEV1 values to normal (healthy) values.
Respiratory speciallists wished to conduct a randomized trial to compare two different drugs in terms of FEV1. The experimenters planned to enlist newly diagnosed patients with moderate to severe asthma and randomize them in equal numbers to the two treatments. Before initiation of the treatments, the subjects underwent spirometry to evaluate their baseline FEV1 values. The subjects would return for spirometry at two month intervals while under treatment. The investigators wanted advice on the number of subjects required and for recommendations on the number of months that subjects should be followed.
Your assignment will be to conduct suitable calculations and provide a short report, explaining your recommendations. We will discuss the problem in next Tuesday's class. Each student is responsible for providing three questions relevant to eliciting the information necessary to provide the requested advice. Please post your questions on the course WIKI by noon on Monday as a basis for class discussion on Tuesday.
I'll play the part of an investigator on Tuesday to answer your questions.
Further questions and discussion will occur in Thurdays lecture,
if required.