USMLE STEP 3 BIOSTATS UPDATED STUDY GUIDE
QUESTIONS AND CORRECT ANSWERS
C
SD 1 SD 68%
2 SD 95% (27% above 1SD, 13.5 on either side)
3 SD 99.7% (5% above 2 SD, 2.5 on either side)
Mean average of a group of number
Median the middle number when arrange in order
Mode the number that appears most
Bell curve features normal- mean = median = mode
positively skewed- mean > median > mode
negatively skewed- mean < median < mode
Accuracy Equivalent to validity; gold standard test; if something is true, it is accurate.
Reliability Reproducible, not necessarily accurate (true)
SEM Measures tightness of grouped dataset (smaller = more precise data)
Z score Measure of where a score lies on normal distribution curve in SDs;
Confidence Interval Probability that 95% of study population mean falls repetitively within given
lower and upper boundaries; (e.g. interval of 80 - 88 for SD=2 for a mean of
84); overlapping ranges a/w statistical significance between groups; wider
ranges a/w smaller sample size
Correlation coefficient (r) Level of connection or correlation b/n two groups
+1 positive correlation
-1 inverse correlation
0 no correlation
T test Assess different groups of data between two means; paired - one
measurement variable and two sample groups; unpaired - two measurement
variables and one sample group
, ANOVA Assess different groups of data between 3 or more groups not normal or bell-
shaped
Chi square test Null hypothesis testing method; checks if observed frequencies in one or more
data categories match expected frequencies.
Radomized control trial Most accurate test, prospective study, avoids selection bias
Cohort Study Determines exposure affecting disease-free groups with similar characteristics
(e.g. age, sex) over time; similar to clinical (experimental) trial
Case Report Aka retrospective study; provides description of individual or a group of
unusual patient cases w/same diagnosis.
Prevalence Survey (cross-sectional) looks at prevalence of disease and risk factors at moment in time
Relative Risk Aka risk ratio; Measures number of cases/number of subjects; used to predict
common disease affecting larger populations; expresses as value greater than
1.0
Odds Ratio Measure of chances an event will occur/chances it will not; mc used in case-
control, retrospective, and multivariate studies, and where smaller population
samples are studied (i.e. low disease risk)
Berkson Bias Hospitalized patients used for study rather than general populations; bias error
mc solved by random selection
Harthorne effect Subject's knowledge of being studied leading to self-modify behavior; mc
solved by double-blind studies and placebo controls
Lead time bias Occurs when test diagnoses disease earlier than another test; produces an
overestimation of disease survival time.
p value Value describing the likelihood test data would have occurred by random
chance (i.e. that the null hypothesis is true); < 0.05 = cutoff for statistical
significance.
Type 1 error (alpha) False-positive result; (ex: test says you're infected but you're actually not);
likelihood of this occurring increases with multiple testing points
Type II error (beta) False-negative result; (ex: test says you're not infected but you actually are)
QUESTIONS AND CORRECT ANSWERS
C
SD 1 SD 68%
2 SD 95% (27% above 1SD, 13.5 on either side)
3 SD 99.7% (5% above 2 SD, 2.5 on either side)
Mean average of a group of number
Median the middle number when arrange in order
Mode the number that appears most
Bell curve features normal- mean = median = mode
positively skewed- mean > median > mode
negatively skewed- mean < median < mode
Accuracy Equivalent to validity; gold standard test; if something is true, it is accurate.
Reliability Reproducible, not necessarily accurate (true)
SEM Measures tightness of grouped dataset (smaller = more precise data)
Z score Measure of where a score lies on normal distribution curve in SDs;
Confidence Interval Probability that 95% of study population mean falls repetitively within given
lower and upper boundaries; (e.g. interval of 80 - 88 for SD=2 for a mean of
84); overlapping ranges a/w statistical significance between groups; wider
ranges a/w smaller sample size
Correlation coefficient (r) Level of connection or correlation b/n two groups
+1 positive correlation
-1 inverse correlation
0 no correlation
T test Assess different groups of data between two means; paired - one
measurement variable and two sample groups; unpaired - two measurement
variables and one sample group
, ANOVA Assess different groups of data between 3 or more groups not normal or bell-
shaped
Chi square test Null hypothesis testing method; checks if observed frequencies in one or more
data categories match expected frequencies.
Radomized control trial Most accurate test, prospective study, avoids selection bias
Cohort Study Determines exposure affecting disease-free groups with similar characteristics
(e.g. age, sex) over time; similar to clinical (experimental) trial
Case Report Aka retrospective study; provides description of individual or a group of
unusual patient cases w/same diagnosis.
Prevalence Survey (cross-sectional) looks at prevalence of disease and risk factors at moment in time
Relative Risk Aka risk ratio; Measures number of cases/number of subjects; used to predict
common disease affecting larger populations; expresses as value greater than
1.0
Odds Ratio Measure of chances an event will occur/chances it will not; mc used in case-
control, retrospective, and multivariate studies, and where smaller population
samples are studied (i.e. low disease risk)
Berkson Bias Hospitalized patients used for study rather than general populations; bias error
mc solved by random selection
Harthorne effect Subject's knowledge of being studied leading to self-modify behavior; mc
solved by double-blind studies and placebo controls
Lead time bias Occurs when test diagnoses disease earlier than another test; produces an
overestimation of disease survival time.
p value Value describing the likelihood test data would have occurred by random
chance (i.e. that the null hypothesis is true); < 0.05 = cutoff for statistical
significance.
Type 1 error (alpha) False-positive result; (ex: test says you're infected but you're actually not);
likelihood of this occurring increases with multiple testing points
Type II error (beta) False-negative result; (ex: test says you're not infected but you actually are)
