Lecture 20- Non Independent Samples Flashcards
What is an example of when two samples would not be considered independent?
Family members if genetics might have an effect
Do you compare independent samples the same way as you compare non-independent samples?
What will result is no confidence (a very large window value). Therefore, you cannot (have to take dependance into account)
What data type infers dependance?
Paired data: where individuals have two measures before and after treatment. These act as your two samples.
How do you compare means correctly for matched data?
-Calculate the differences between the paired data points. These
become the new data. (this because with paired data its more important to calculate the difference within individuals than between individuals= a lot more variation and wider resulting interval)
-Then use the confidence interval for a single mean
What is d bar?
just minus the the mean of the second group off the mean of the first group
A nutrition scientist is assessing a weight-loss programme to evaluate its effectiveness. Twenty people were randomly selected to take part.
Both the initial weight and the final weight after four months on the
programme were recorded.
Look at data on lab slide and calculate the confidence interval assuming dependance….
Working on slides but answer=
3.9kg< µd < 8.4kg
Taking into account the paired design, we are 95% confident that the true mean difference in weight lies between 3.9 and 8.4 kg. Since this interval does not include zero and is entirely positive, there is evidence that weight is lower after the diet than before. On average, people are likely to be between 3.9 and 8.4 kg lighter after undergoing the diet programme
What is regression to the mean?
Weight varies naturally over time. It may be the people entering
the weight loss programme do so because their weight is higher than
usual and they wish to lose weight. The loss of weight may reflect
natural variation in weight due to other life factors, rather than
effectiveness of weight loss programme
What is the best way to separate a true effect of the programme from regression to the mean?
RCT= both groups have same motivation (told to lose weight by doctors) but 1 group also have nutrition advice (only difference). Therefore because of the addition of a control group we can tell if the program is truly effective.
