Data analysis 1 : comparing mean and medians Flashcards
Why do we perform statistical analysis?
- to determine whether there is sufficient evidence to accept or reject the experimental hypothesis
- to test if the data is both genuine(was it random?) and meaningful (is effect size biologically or clinically relevant?)
What is a p value
- calculated probability of observing an effect size to see if there was no genuine difference between the groups (ie. if all of the data points for the groups being compared actually came from the same group)
How do you accept or reject the null hypothesis based on p value and alpha value- interpreting p value?
- p values are compared to threshold value ‘alpha’
- p< alpha = the difference is statistically significant (reject null hypothesis)
- p >/= alpha = the difference is not statistically significant (accept the null hypothesis)
What is an alpha value?
the threshold value that we use to compare p value to so we can see if the results are significant or not.
Why is the most commonly used alpha value 0.05?
- 1920s Fisher thought it is fairly unlikely thing to happen 1/20 chance (5%).
- value of convention
Choice of alpha value(0.05 or 0.001) considers …?
- nature of data of the study and relative consequences of type 1 and type 2 errors
- small alpha is more likely to result in false negative(type 2 error)
- large alpha is more likely to result in false positive (type 1 error).
What is type 1 error?
-false positive
What is type 2 error?
- false negative
What does type of statistical test to use when analysing if data is statistically significant depend on? (why not use t- test all the time?
- if the data adheres to normal distribution (parametric)?
- if the data is from independent samples(paired or unpaired?)
- on how many groups or differences are being compared?
What is type of statistical test do you use for analysing parametric data from independent samples (unpaired data) and why?
- unpaired t- test
- the method by which the p-value is calculated in t test assumes the data adheres to a normal distribution(parametric)
- t test is only appropriate in analysing parametric data.
What determines if p value is large or small?
large : P>alpha
- small sample size
- small difference in group mean
- large standard deviation
small sample size:
- large sample size
- large difference in group mean
- small standard deviation
What is two tailed test?
- parameter that describes whether the analysis is concerned with assessing the statistical significance of potential positive and negative difference between one group and another (two tailed)
What is one tailed test?
- parameter that describes whether the analysis is only concerned with the significance of a difference in one specific ‘direction’
When do you use one tail t test and two tailed t test for?
1 tail: weight of group 1 is significantly different to the mean weight of samples in group 2
2 tail : the mean weight of samples in group 1 is significantly greater than the mean weight of samples in group 2.
Why can’t t test be used to analyse non-parametric data?
- parametric methods generate p values by assuming a specific relationship between the variation of data points to the mean and the probability of observing the, For non parametric data, the relationship doesn’t exist so any test that assume the relationship exists will generate inaccurate p value.
