Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
66 Cards in this Set
- Front
- Back
Selection Bias
|
Non random assignment to study group
|
|
Measurement Bias
|
Measurement is dissimilar among groups
|
|
Confounding Bias
|
Occurs with 2 closely related factors. The effect of 1 factor distorts or confuses the effect of the other.
|
|
What two biases can co exist
|
Selection bias and confounding bias
|
|
Chance
|
Random variation with results likely to be abnormally high or low
|
|
Validity
|
How true are the results of the population studied ? was the study correctly done without bias and chance
|
|
Generalizability
|
Also called external validity. How much do the results apply to other settings
|
|
Reliability
|
The extent to which the repeated measurements are similar. Compare to precision, reproducibility - where the crosses hit the target at the same spot
|
|
Validity
|
Does the measurement represent what it is supposed to. eg. accuracy. All the crosses hit the target
|
|
Interval data
|
Known interval between two successive values. Two types - continuous and integer. Most physiological data is interval
|
|
Nominal data
|
occurs in categories without order. Also called categorical. if 2 options called dichotomous
|
|
Ordinal data
|
Inherent ordering without standard intervals ebtween measurements
|
|
Measure HCT of the class. What are the frequently occurring values?
How much variability is there in the class ? |
Central tendency
Dispersion |
|
Central tendency
Mean Median Mode |
Mean - sum of all values/ number of measurements
Median - number of observations above the value equals the number below Mode - most frequently occurring value |
|
Dispersion
Range Standard deviation Percentile |
lowest to highest
absolute value of average differences of individual values from the mean. 1 standard dev is 68% and 2 is 95.4% and 3 is 99.72% |
|
When mean = median
|
Then mean and SD appropriate - NORMALLY DISTRIBUTED
|
|
WHen mean does not equal median
|
Then NOT NORMALLY DISTRIBUTED. Median and percentiles are appropriate
|
|
How is significance of a value estimated?
|
Using the P value
|
|
What is the P value
|
It measures how likely a particular difference between groups is to be due to chance
|
|
If the difference is 1/20 times
|
Then this is large enough for it to be considered unlikely to be due to chance.
this P value would be .05 or 5% |
|
Characteristics of the T test
|
Most commonly used statistical test in literature
Data must be approximately normally distributed and only two groups can be compared Calculated using means, standard deviations and number of subjects in groups |
|
what is a t test called when each person serves as their own comparision ?
|
A paired T test
|
|
How is the t test when used for multiple comparisions ?
|
It will tell you which groups are different from each other, but it might not tell you which one might be MOST different
|
|
What is the bonferroni correction ?
|
It is the correction for increased chance of error when doing multiple T tests. Divide .05 by the number of test performed and this will represent the actual value for statistical significance
|
|
If you are comparing rates or percentages which test would you use ?
|
Chi Square test
|
|
What is confidence interval ?
|
CI tells you there is a 95% chance that the CI includes the TRUE difference in mortality
|
|
Why is confidence interval more significant than the p value ?
|
It provides more information than p values alone because the reader can view the range of possible true values
It also gives a perspective concerning sample size and power |
|
When data is not normally distributed, what tests should they use ?
|
Non parametric tests - Authors chose this test on purpose, otherwise they rather use a t test.
|
|
Rank sum test
|
Add up the ranks in each group. If the range is 6-18. How unusual is 9.
|
|
What test would you use for 2 groups, 3 groups and a before and after test if u were using a Interval scale
|
t test
Analysis of variance Paired t test |
|
What test would you use for 2 groups, 3 groups and a before and after test if u were using a NOMINAL SCALE
|
Chi square
Chi Square McNemars test |
|
What test would you use for 2 groups, 3 groups and a before and after test if u were using a ORDINAL SCALE
|
Rank sum test
Kruskall Wallace Signed Rank test |
|
Intervals
|
# Nominal Data
* classification data, e.g. m/f * no ordering, e.g. it makes no sense to state that M > F * arbitrary labels, e.g., m/f, 0/1, etc # Ordinal Data * ordered but differences between values are not important * e.g., political parties on left to right spectrum given labels 0, 1, 2 * e.g., Likert scales, rank on a scale of 1..5 your degree of satisfaction * e.g., restaurant ratings # Interval Data * ordered, constant scale, but no natural zero * differences make sense, but ratios do not (e.g., 30°-20°=20°-10°, but 20°/10° is not twice as hot! * e.g., temperature (C,F), dates |
|
Prevalence ?
|
What percentage of people at one given time have this disease
|
|
What is a gold standard ?
|
Usually an invasive or expensive test that provides a high degree of accuracy and is assumed to be correct.
|
|
HOw is prevalence calculated ?
|
A+C/EA
|
|
What is sensitivity ?
|
Percentage of diseased patients that have a positive test
A/A+C |
|
What is specificity ?
|
Percentage of non-diseased pts that tested negative
D/D+B |
|
What determines cutoff points for sensitivity and specificity ?
|
Cut off levels
|
|
What affects predictive values and not specificity or sensitivity
|
The prevalence of the disease affects the positive or negative predictive values
|
|
To rule out a disease do you want high what ?
|
Sensitivity
|
|
To confirm the presence of a disease, you want a high ?
|
Specificity
|
|
What is a Likelihood ratio ?
|
How much more likely is a test to be positive among those with disease as opposed to those without disease
Test with a high specificty and a high sensitivity will have a high likelihood ratio |
|
If the disease causes a serious outcome, you want a high ?
If the treatment is invasive and possibly dangerous you want a high ? |
Sensitivty and Specificty
|
|
Problems in the evaluation of diagnostic tests
|
What is Normal
Spectrum of patients Bias Chance |
|
SpPIn
|
SnNout
|
|
What is prevalence and incidence ?
|
Prevalence is the proportion of the population that have a particular characteristic at one point in time
Incidence describes the proportion of a population initially free of disease that develop the disease over a specific period of time |
|
Case control study
|
Doesnt take into account incidence. Women with cancer who took estrogens.
Women without cancer who took estrogens |
|
Cohort study
|
This tests incidence
Women without cancer taking estrogens Women without cancer not taking estrogens. Looking for the development of cancer |
|
Measuring a Cohort study using Relative risk
|
Incidence in the exposed/Incidence in the unexposed
RR= A/A+B divided by C/C+D |
|
Case control using Odds ration
|
Rate of exposure in the cases/Rate of exposure in the controls
OR = AXD/BXC |
|
Concurrent cohort
|
Those who are exposed at the start vs unexposed
|
|
Historical Cohort
|
Requires records of patients, or recall leading to recall bias. Based on incidence of disease, they look back at records
|
|
Case/Control
|
Start at presence or absence of disease, and check if they took estrogens or not
|
|
which study always begins with exposure and which one with disease ?
|
Cohort-exposure
Case control - disease |
|
Which one has incidence ?
|
Cohort
|
|
Understand absolute vs relative risks
|
These are not the same
|
|
Accuracy of survival curves better on which side ?
|
Left side
|
|
Advantage of these survival curves
|
They can provide information about patients entering studies at varying points in time and increases the efficiency of the studies
|
|
Higher the Confidence intervals, what happens to the accuracy ?
|
Lower the confidence. indicating probably a lower sample. higher the sample, more narrower is the confidence interval
|
|
Advantages of Cohort studies
|
Only way of establishing incidence directly
Follows logic of clinical questions Exposure can be elicited without BIAS Can assess relationship between exposure and many diseases |
|
Disad
|
Inefficient
Expensive Long time to get results Study limited to only those variables measured at start |
|
Adv of case control studies
|
Excellent for rare disease
Quick Cheap Can evaluate multiple risk factors |
|
Disadvantages
|
Bias
Selecting cases Selecting controls Measuring exposure |
|
Odds ratio in case control
|
Odds ration measures the risk of exposure in one group vs other
Investigator determines groups especially controls Anything OTHER than the measured variable that can affect the rate of exposure will cause BIAS |
|
Measurement in Case control
|
The presence of the outcome directly affects the exposure
The presence of the outcome affects the subjects recollection of the exposure - Recall bias The presence of the outocme affects the measurement of recording of the exposure |