CNSL 503: A GUIDE TO A+. QUESTIONS AND ANSWERS
FORMAT.LATEST UPDATE
When the prevalence of a condition/disease is low in a population, how are the positive and negative
predictive values affected?
lower PPV & higher NPV
Can positive and negative predictive values be generalized for a specific test?
No, because the prevalence of disease is most likely not the same in all populations
Can sensitivity and specificity be generalized for a specific test?
Yes, because the prevalence of the disease doesn't affect the values
Sensitivity (SnOUT)
ability of test to obtain positive result when one is truly positive, so it is very good at picking up positive
results. Therefore when it is negative, you can rule OUT a diagnosis
Sensitivity (Sn) formula
Sn = TP/(TP+FP)
true positive results / all actual positive people
Specificity (SpIN)
ability of test to obtain negative result when one is truly negative, so it is very good at picking up
negative results. Therefore, when the result is positive, you can rule IN a diagnosis.
Specificity (Sp) formula
Sp = TN/(TN+FN)
true negative results/ all actual negative people
A test with high _____________ that when negative helps rule OUT a condition.
A test with high sensitivity that when negative helps rule OUT a condition.
A test with high _________________ that when positive helps rule IN a condition.
A test with high specificity that when positive helps rule IN a condition.
Highly sensitive tests may commit what type of error?
false-positives = type 1
Highly specific tests may commit what type of error?
false-negatives = type 2
, Can likelihood ratios be generalized for a specific test?
Yes, because like sensitivity and specificity they do not depend on the prevalence of a condition in a
population
Likelihood ratios
combine sensitivity and specificity to indicate the shift of probability, given a specific test result
Positive Likelihood Ratio (LR+)
indicates an increase in odds of having the condition if one tested positive
Positive Likelihood Ratio (LR+) formula
LR+ = Sn/(1-Sp)
-ratio of true positive results to false-positive results
Negative Likelihood Ratio (LR-)
indicates a decrease in odds of having a condition if one tested negative
Negative Likelihood Ratio (LR-) formula
LR- = (1-Sn)/Sp
ratio of false-negative results to true negative results
The larger the positive likelihood ratio, the more or less likely you are to have the condition?
The larger the positive likelihood ratio, the more likely you are to have the condition?
The smaller the negative likelihood ratio, the more or less likely you are to have the condition?
The smaller the negative likelihood ratio, the less likely you are to have the condition?
Interpreting Likelihood Ratios Table
What does a likelihood ratio of 1 mean?
a 50/50 chance of increasing or decreasing the probability of a diagnosis
- you do NOT want to use a special test if it has a LR = 1, it won't help you
Correlation vs Causation
- Correlation = relationship between two variables
- Causation = one variable producing/predicting an effect in another variable
correlation does not equal causation
Multiple Linear Regression
FORMAT.LATEST UPDATE
When the prevalence of a condition/disease is low in a population, how are the positive and negative
predictive values affected?
lower PPV & higher NPV
Can positive and negative predictive values be generalized for a specific test?
No, because the prevalence of disease is most likely not the same in all populations
Can sensitivity and specificity be generalized for a specific test?
Yes, because the prevalence of the disease doesn't affect the values
Sensitivity (SnOUT)
ability of test to obtain positive result when one is truly positive, so it is very good at picking up positive
results. Therefore when it is negative, you can rule OUT a diagnosis
Sensitivity (Sn) formula
Sn = TP/(TP+FP)
true positive results / all actual positive people
Specificity (SpIN)
ability of test to obtain negative result when one is truly negative, so it is very good at picking up
negative results. Therefore, when the result is positive, you can rule IN a diagnosis.
Specificity (Sp) formula
Sp = TN/(TN+FN)
true negative results/ all actual negative people
A test with high _____________ that when negative helps rule OUT a condition.
A test with high sensitivity that when negative helps rule OUT a condition.
A test with high _________________ that when positive helps rule IN a condition.
A test with high specificity that when positive helps rule IN a condition.
Highly sensitive tests may commit what type of error?
false-positives = type 1
Highly specific tests may commit what type of error?
false-negatives = type 2
, Can likelihood ratios be generalized for a specific test?
Yes, because like sensitivity and specificity they do not depend on the prevalence of a condition in a
population
Likelihood ratios
combine sensitivity and specificity to indicate the shift of probability, given a specific test result
Positive Likelihood Ratio (LR+)
indicates an increase in odds of having the condition if one tested positive
Positive Likelihood Ratio (LR+) formula
LR+ = Sn/(1-Sp)
-ratio of true positive results to false-positive results
Negative Likelihood Ratio (LR-)
indicates a decrease in odds of having a condition if one tested negative
Negative Likelihood Ratio (LR-) formula
LR- = (1-Sn)/Sp
ratio of false-negative results to true negative results
The larger the positive likelihood ratio, the more or less likely you are to have the condition?
The larger the positive likelihood ratio, the more likely you are to have the condition?
The smaller the negative likelihood ratio, the more or less likely you are to have the condition?
The smaller the negative likelihood ratio, the less likely you are to have the condition?
Interpreting Likelihood Ratios Table
What does a likelihood ratio of 1 mean?
a 50/50 chance of increasing or decreasing the probability of a diagnosis
- you do NOT want to use a special test if it has a LR = 1, it won't help you
Correlation vs Causation
- Correlation = relationship between two variables
- Causation = one variable producing/predicting an effect in another variable
correlation does not equal causation
Multiple Linear Regression
