Exam 1

Question 0

Sensitivity

Answer 0

Probability of a positive test for individuals with a given disease

Question 1

Specificity

Answer 1

Probability of a negative test for individuals without a given disease. Is intrinsic property of test and does not depend on prevalence. Is intrinsic property of test and does not depend on prevalence.

Question 2

Cut off point

Answer 2

The threshold for where you decide a screening test is positive. Defines the disease. Yields different specificity and sensitivity results depending where you put it.

Yes/no (has or does not have disease) is binary but most variables like blood sugar are continuous.

Setting the cut off point will treat the lower scores as normal.

Question 3

Event

Answer 3

Measurable outcome or case

Question 4

Sample vs population

Answer 4

Sample average is with x bar and sample standard deviation is “s”. Pop average is mu and pop standard deviation is sigma.

Question 5

Negative predictive value

Answer 5

Proportion of screening tests that are true negatives. Describes the performance of a screening test. Is not inherent to the test because it also depends on prevalence. Can be derived using bayes theorem. (# true negatives / all negatives)

Question 6

Positive predictive value

Answer 6

Proportion of screening tests that are true positives. Describes the performance of a screening test. Is not inherent to the test because it also depends on prevalence. Can be derived using bayes theorem. =# true positives / all positives.

Lots of false positives in large low prevalence population unless specificity is extremely good.

Question 7

Median

Answer 7

50th percentile. Survives outliers. Robust measure. Comparing median to mean can show skewness

Question 8

IQR

Answer 8

Inter quartile range 25th-75th percentile. Calculate by finding the difference between the middle point between the lowest data point and the median and the highest data point and the median

Question 9

Percentile

Answer 9

You can also index by multiplying the number of points in the set by the percentile (15 scores*.9=13.5 which means the 90th percentile would be the 14th highest score)

On excel: =percentile (A2:A19, .9)

Question 10

Variance

Answer 10

Measures how far a set of numbers is spread out. Always non negative. Smaller means the data points tend to be very close to the mean (expected value)

Question 11

Standard deviation

Answer 11

34/68(1)
95/47.5(2)
99.7(3)

Measures spread and summarizes data

Question 12

Skewness

Answer 12

A tail to the left or right of data. Left slew indicates something that tends to occur later or higher, for example mortality rate. Right skew indicates something that tends to occur earlier or lower, like income.

Question 13

Boxplot

Answer 13

Summarizes data around mean. Whiskers extend beyond 1st and 3rd percentile. Outliers indicated by dots.

Question 14

Histogram

Answer 14

Data collected in bins. All data points should fit in a bin. Artificial categories or ranges are created.

Question 15

Conditional probability

Answer 15

A|B does not = B|A.

Bayes rule lets you go back and forth

Question 16

Union

Answer 16

A or B

Question 17

Intersection

Answer 17

A and B

Question 18

Empirical rule

Answer 18

Use z score to determine how many standard deviations a certain case is from the mean in a normal distribution.