(incomplete) F - Statistical Concepts I-II Flashcards
Introduce contingency table analysis and hypothesis testing for 2x2 tables. (Chi squared test)
Null hypothesis: two groups are the same
Chi squared test: allows you to tell if two groups are the same or different
- -Compute expected cell frequencies assuming null hypothesis is true by calculating the overall frequency
- -Apply the proportion (overall frequency) to the two groups
- -Calculate: sum of [(O-E)^2]/E
- —Larger value = greater evidence of association
- —Compute p value with a table
- —For 2x2 table (four cells, four sums, 1 degree of freedom), if value is greater than 3.84, then p is less than 0.05
Conclusion if significant: “the probability of observing this level of association, or a more extreme level of association, if the two groups really had the same proportion is small (very unlikely”
Introduce the independent samples t-test for comparing the means of two independent samples.
Group effect (difference) = x1 - x2
Standard error of difference = sqrt([(SD1)^2)/n1] +[(SD2)^2)/n2])
t statistic = group effect/standard error of difference
- -p value = sum of total area beyond +t and -t
- -Significant if t is at least 2.0 and sample size is at least 50 in each group
Introduce the paired-samples t-test for comparing the means of two paired samples.
Evaluates a continuous FINISH??
t = mean of differences divided by the result of the standard deviation of the differences divided by the square root of n t = d/(SD/sqrt(n))
Introduce survival analysis methods.
a
How can the size of an effect be estimated? Is p value useful? State the null value for each of the three estimations.
p value gives no estimate of size of effect
(A and B are proportions)
Risk difference: A-B (null value = 0)
Relative risk: A/B (null value = 1)
Odds ratio: [A/(1-A)]/[B/(1-B)] (null value = 1)
–Odds ratio will be similar to relative risk if event is very rare
Do significant results indicate causal effects?
NO. There could be confounding.
