Epidemiology Week 3 Flashcards Preview

Fundamentals > Epidemiology Week 3 > Flashcards

Flashcards in Epidemiology Week 3 Deck (22):
1

Describe the nominal measurement scale

◾data that can be named and put into categories
◾categories are mutually exclusive and unordered

2

Describe the ordinal measurement scale

◾named data
◾categories are mutually exclusive and ordered

3

Describe the count/discrete measurement scale

◾numeric/measured data
◾specific values, integers

4

Describe the continuous measurement scale

◾numeric/measured data
◾data that can take on any value

5

What is central tendency?

measures of location (ex. mean, median)

6

Give four measures of central tendency

◦Mean - arithmetic average
◦Median - middle value of sorted data, less influenced by outliers
◦Mode - most frequent value
◦Geometric mean - nth root of the product of the data

7

What is variability?

measures of spread (ex. range, std dev)

8

Give different measures of variability

◦Range - max value - min value
◦Interquartile range - 75th percentile - 25th percentile
◾Captures central 50% of the data
◦Variance - average squared deviation from the mean
◦Standard deviation - square root of variance
◦Coefficient of variation - standard deviation/mean

9

What are the major sources of variation?

◦Biologic variation
◾Between-individual
◾Within-individual
◦Measurement variation
◾Between-observer
◾Within-observer
◦Instrument or analytical error

10

What is reliability?

- Precision
- Random error

11

What is validity?

- Accuracy
- Systematic error

12

How is data displayed graphically?

- Histograms
- Box charts
- Bar charts/Pie charts
- Distribution curves

13

Give the properties of normal distribution curves

◾Bell shaped
◾Total area under curve = 1
◾Extends to infinity in both directions
◾Probability corresponds to area under the curve
•68% within +/- 1 standard dev of mean
•95% within +/- 2 standard devs of mean
•99% within +/- 3 standard devs of mean
◾Only need to know mean and standard dev

14

What is sensitivity:

Proportion of those with disease who also have a positive test result
Sensitive test means TP/(TP + FN) close to 100%
few false negatives
most negative results are true

SNOUT (Sensitive test with negative result rules out disease)
Sensitivity and negativity both have Ns in them

15

Specificity

Proportion of those without the disease who give a negative test result
Specific test means TN/(TN+FP) close to 100%
gives few false positives
most positive results are true
Patients with positive results likely have the disease
SPIN = Specific test with Positive results rules in disease
Specificity and positive both have Ps in them

16

False positive and false negative

If FN is worse than FP use a sensitive test
If FP worse than FN, use a specific test

17

Positive predictive value

Calculate the probability of the patient having the disease given a positive test result
TP/(TP+FP)
prevalence is positively associated with +PV (prevalence increases, +PV increases)

18

Negative predictive value

Calculate the probability of the patient not having the disease given a negative test result
TN/(FN+TN)
prevalence is negatively associated with -PV (prevalence increases, -PV decreases)

19

Use properties of the normal (Gaussian) distribution and the standard normal (Z) distribution to estimate probabilities of clinical events:

normal (Gassian) distribution: describes frequency distribution of repeated measurements of the same physical object by the same instrument
standard normal distribution:
- curve is symmetrical and bell-shaped
- two-thirds of observations fall within 1 standard deviation of mean
- 95% of observations fall within 1 standard deviation of the mean

20

Methods to determine reference intervals: normal distribution method

mean +/- 2x standard deviation
68% within +/1 1 standard dev of the mean
95% within +/- 2 standard dev of the mean
99% within +/- 3 standard dev of the mean

21

Transformation method:

transform values to make data more like a normal distribution using logs
use the normal distribution method

22

Percentile method:

assumes no distributional form
works for any distribution (nonparametric)
sample data should be representative of reference population
sample size should be large enough so estimates of range will be reliable

Decks in Fundamentals Class (66):