Analysing variance and analysis of repeated measurements Flashcards Preview

Stats intercal > Analysing variance and analysis of repeated measurements > Flashcards

Flashcards in Analysing variance and analysis of repeated measurements Deck (11)
Loading flashcards...

What is the principle behind analysis of variance?

to partition the total variability of a set of data into components due to different sources of variation
systematic differences between the group means and
the variations between individuals within each group


What is the null hypothesis in one-way analysis of variance?

The is no difference been the groups and the test compares the observed variation between the groups with that elected from the observed variability between subjects


What are the assumptions underlying ANOVA?

the samples come from normally distributed populations with the same standard deviation
normality of residuals


What is the problem of making multiple comparisons?

gives a high probability of a finding a significant difference just by chance


What is the bonferroni method?

a method of manual multiple testing
it is fine for up to 5 comparisons but is highly conservative
it adjusts the p value to account for multiple testing


Describe Tukey's HSD

most sensible choice if you wish to consider all comparisons


Describe Dunnett's

most sensible choice when you are only interested in comparing one group (e.g control) to each of the other groups. More powerful than Tukey's in this situation


Describe Fisher's LSD

not generally recommended. does not take multiple comparisons into consideration


What is different about the analysis of repeated measurements?

observations on the same subject just have some relationship between them, and are very likely to be correlated to each other. Data points are not independent


What are the problems with point by point analysis of repeated measurements?

ignores the correlation between successive observations from the same individual
may conceal important features of the data
causes a multiple testing problem
ignores the continuous nature of the underlying process


Give examples of summary measures that can be used for each subject's series of measurements

maximum or minimum values
time to maximum
area under the curve
slope of straight line fir
last measure minus first measure