Stats PPT 3 Flashcards
Name some types of scientific/medical studies: (4 examples)
Randomized controlled trial
Prospective trial
Retrospective study
Cross-sectional study (prevalence study)
What does a randomized, controlled trial do?
Randomized: no pattern to who is distributed into one group vs another. Eliminates bias.
What does blinding do?
What does double blinding do?
Blinding = patients don’t know.
Double blinding = Both managers and patients don’t know
What makes a trial prospective?
You establish what you are looking at before you actually study people.
When is a retrospective study more useful?
When studying a disease that is very rare/uncommon. You can’t be certain it will turn up in a prospective trial.
What is a retrospective study?
When you take data that has already been generated and look back at it.
How do you randomize?
Sometimes they use the numbers in Pi.
What distinguishes an experimental study from an observational one?
Experimental = actively doing something Observational = just watching and recording
What is a case series?
Kind of like a smaller retrospective.
What is a meta-analysis?
A look at lots of complicated studies, collectively.
How does a multi-stage therapeutic trial minimize harm?
You start small. Check carefully. Then increase and follow up.
What are some methods of data display?
Error bars, box-and-whisker plot,
pie chart, scatter plot
What do error bars do?
Indicate the range of probable error.
Show average and variability for each parameter.
What if the error bars overlap?
Might diminish significance of differences.
What information does a box-and-whisker plot give you?
Total range up and down, median (in the middle), quartiles.
What kind of display tells you about the course of follow-up for (ex: bypass graft) patients?
Life table
What does linear regression tell you?
The probability of correlation/trend. It generates an r value.
Is it bad to see a negative correlation?
No, just an inverse relationship.
Can you glance at a scattergram and tell roughly whether there is correlation?
Maybe
What does the r value tell you?
r = regression. The likelihood of correlation.
What’s a good r value?
> 0.8
What’s a perfect r value?
Strongest r value = 1.00
What r value shows absolutely no correlation?
Weakest r value = 0.00. Totally random.
What things might make you more skeptical about a study? (7 examples listed)
Where did the funding come from? Is there data missing? Was the patient dropped? Is the test repeatable? Was the study published? Was it peer reviewed? Was there dubious sampling?
