ch 9: Student T-Test Flashcards
what is the student t test
used when looking for a differene in the mean value of an interval level or ratio level variable
what 3 things do you need to use before you can apply the student t test
o What is the level of measurement for the outcome variable (need interval or ratio)
o Are there 2 samples
o Aret he samples independent and random
what level of meas is nec in order o use the student t test
interval or ratio
in this example are the samples independent?
looking at difference in amount of epi in injectors (ratio level) so you would look at what kind of samples you have so in this ex. Injectors from acme and a sample of injectors from epi and looking at diference in mean amount of epi in both
. In example neither of them affected one another so they are independent samples.
when might you have dependent samples
is this a dependent sample: you want to meas duration of stay at dif hospitals and focus on 10 pts at one hospital with specific Dx then go to another hospital and look for pts with same Dx
- if you sample the same group but at a diff point in time
- • Can also have dep samples when you match sample char. Ie.) Measuring duration of stay at different hospitals and you decide to choose 10 pt with same dx at each hospital. They are diff samples but share char.
dependent sample defn
paired or relate groups or the same sample at a different tim
noninferiority trial
a trial used to show that a ew Tx is no worse than an old procedure (may use a one-tailed test)
what is a sampling error
error that occurs due to randomization AND CHAnce
to determine if the diference is stat signif what must you decide on
o Level of risk you are taking that will incorrectly reject the null hypothesis (alpha). Ie. deciding there is a 5% chance so setting alpha at 0.05
o The power (chance of finding a diff if it actually exists) at 0.80 or 80%
o Whether you are coudcint g a one tailed or two tailed test.
if set alpha at 0.05 what are you saying ain r/t errors
theres 5% chance youll make type one error
what is a one tailed test vs a two tailed test
Meaning whether you are looking for a diff in either direction (two tailed) or you have hypothesized that the diff is in a direction (one tailed). Ie.) looking for a direction between epi levels in 2 groups that is a two tailed student t-test
is using a one tailed test generally supported
what is the assumption
• Is controversial and should be avoided. If youre wanting to do a directional null hypothesis that is one-tailed test
-there is no chance of a change in one of the two possible directions for a possible variable. Therefore, lpha which is usually split between tw tails has its prob shifted to one tail or one side
what is the degrees of freedom for student t tests
equal to the sample size for both groups minus two (this is same as taking the degrees of freedom for each group (or sample size for the group minus one) and adding them together
when you have 10 epi injectors and meas their epinephrine level, you have 10 values but once you calculate the mean level of epinephrine in each injector how many values are free to be unknown?
Now once you know the amount in eachof the 9 and know the mean is the value for the last injector pen free to vary?
what does this mean for degrees of freedom?
9
no the amount is no longer free to vary
when you calculate the mean you lose a degree of freedom
so basically for degrees of freedom
take your total sample size (both groups) minus two (one from each group) and you know the degrees of freedom for the test which is all you need to then find the corresponding p value on the t distribution table
.
since the t test has a few versions what questions do you need to ask before performing them
o How many tails of the statistical distribution do you want to test?
o Is your data paired?
o Can you assume equal variances in the two different groups?
if you were studying pts who were in a crossover clinical trial and at one point receive a placebo then later receive a drug (or vice versa) what kind of analysis would you have to do
paired analysis
-the pts are paired instead of being independent
..• If the samples are dependent, the groups are matched on an attribute or may be the same group measured at a different time; in either case they are related to each other.
..