Biostats Flashcards
Categorical data
Data fits into one of several categories
Ex: gender
No order (male = female) No magnitude (size) No decimals used
Integer data
Whole number that can be positive negative or zero
Ex: Number of blood cultures
Ordered (1<2)
Magnitude differences (size difference)
No decimals
Ordinal data
Data that has natural, ordered categories
Ex: ROP stage
Ordered (1<2)
No magnitude difference
No decimals used
Continuous data
Data whose value can be obtained by measuring
Ex: birth weight
Ordered
Magnitude difference
Decimals used
Mean
Average
Sum of all values / N
Most sensitive to outliers
Median
Middle value
Mode
Most common value
Peak
Range
Highest - lowest
Outliers can skew this
Interquartile range
75th % - 25th %
(middle 50%)
Eliminates outliers
Mean deviation
(Sum |X-mean|) / N
Hard to manipulate
Standard deviation
Average distance of each value from the middle
What kind of error do you have if you reject the null hypothesis when it is true?
Type 1 (alpha) error
Concluding there is a difference between groups when there is not
False positive
What kind of error do you have when you accept the null hypothesis that is false?
Type 2 (beta) error
Concluding there is no difference between groups when there is
False negative
What statistical test should you use to compare 2 separate groups of continuous data?
Unpaired t-test
What statistical test should you use to compare 2 paired groups of continuous data?
Paired t-test
What kind of statistical test should you use to compare three or more groups of continuous data?
Analysis of variance (ANOVA)
Multiple regression
What kind of statistical test should you use to compare 2 separate groups of categorical data?
Chi-square test
What kind of statistical analysis should you use to compare 2 paired groups of categorical data?
McNemar’s
What kind of statistical test should you use to compare three or more groups of categorical data?
Chi-square
Logistic regression
Nonparametric test for two separategroups of continuous data?
Man Whitney
Nonparametric test for two paired groups of continuous data?
Wilcoxon
Nonparametric test for three or more unmatched groups of continuous data?
Kruskal Wallis
The P value is the same as what type of error?
Type 1 or alpha error
What does the P value tell you?
The probability of observing a difference at least this large by chance alone
What does the confidence interval tell you?
Based on the sample standard deviation and its size you are 95% confident that the limits cover the true value for the population mean
What is a Case series?
Choosing cases with an exposure and following them to see the outcome
What is a case control study?
Finding patients with an outcome (and control patients) and looking back to see if either group was exposed
What is a cohort study?
Finding patients with or without an exposure/disease and following forward to see what their outcome is

Less biased than case control studies
What is a Randomized controlled trial?
Randomly chosen cases and controls (exposed/non-exposed) followed to see their outcome
Which type of study is not good for rare outcomes?
Cohort study
What does relative risk reduction tell you?
The risk of an adverse outcome in the experimental group is reduced by this proportion relative to controls
What does an odds ratio tell you?
Odds of adverse outcome in experimental group reduced by this proportion relative to controls
What does absolute risk reduction tell you?
Risk of adverse outcome in the experimental group is reduced by this absolute percentage
Odds ratio = relative risk when the prevalence of disease is ___
Low
Odds ratio will always ____ The effect size compared to relative risk
Overestimate
What does sensitivity mean?
How many patients with the disease will have a positive test
What does specificity mean?
How many patients without the disease will have a negative test
What does positive predictive value mean?
How many patients with a positive test actually have the disease
Affected by disease prevalence
What does negative predictive value mean?
How many patients with a negative test actually don’t have the disease
Affected by disease Prevalence
Run chart rules
1. Shift: >6 points above or below the median
2. Trend: >5 points upper down in a row
3. Two few or too many runs
4. Astronomical Datapoint
What does a SMART aim need to be?
Specific Measurable Achievable Relevant Time-limited
What is a process measure?
A measure tracking whether you were following protocols
What is an outcome measure?
Whether you meet your goal
What is a balancing measure?
Looking at possible trade-offs
Standard deviations
1SD 68%
2SD 95%
3SD 99.8%
Ways to increase power of a trial
Increase sample size
Increase effect size
Increasing type one error rate
Decreasing standard deviation of the outcome
Which is more important, the absolute difference or the relative difference?
Absolute difference
