Community Medicine

Kish grid

PSM / COMMUNITY MEDICINE by Dr Abhishek Jaiswal is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Based on a work at learnpsm@blogspot.com.
Permissions beyond the scope of this license may be available at jaiswal.fph@gmail.com.

Likelihood Ratio

•Likelihood ratio +: How much the increase in the probability of having the disease if the test result is positive

LR+ = Sn/(1-Sp)

•Likelihood ratio-: How much decrease in the probability of having the disease if the test result is negative

LR- = (1-Sn)/Sp

Cronbachs Alpha

•Cronbach’s Alpha:

It is a measure of internal consistency of a test/scale

•Measure that the scale is consistent measure of a concept

•Computed by correlating the score for each item of the scale with the total score for each observation and then comparing

that to the variance of all individual item scores

•It is a f(x) of number of items in a test, average covariance between pair of items, and variance of total score

•Minimum recommended value 0.65-0.8

Cohen Kappa

•Kappa:

It expresses the extent to which the observed agreement exceeds that which would be expected by chance

alone, relative to the maximum that the observer could hope to improve their agreement

•Quantifies the extent to which the observed agreement that the observer achieved exceeds that which would be

expected by chance alone

ROC Curve

ROC Curve: it is a plot that displays full picture of

trade-off between the sensitivity and

specificity across a series of cut-off points

•Use:

•Evaluation of discriminating ability of test

•For finding optimal cut-off point

•Comparing efficacy of two or more tests

•Comparing two or more observer measuring same test

•Area under the curve (AUC): Measure of

validity of the test

Diagnostic Odds ratio

Diagnostic odds ratio:

It relates odds that the test is positive in diseased population to the

odds that the test is positive in healthy population

>=50 : Very strong validity of test

20 to <50 : Strong

10 to <20 : Fair

<10 : Weak

Youden's Index

•Youden’s index:

It is the maximum vertical distance from the line of equality to the

point on ROC curve

•A measure of summary of ROC curve

•It cuts the ROC curve at the point (optimal cut-off point) that

optimizes the differentiating ability when equal weight is given to

sensitivity and specificity

Reliability and Validity

•Reliability:

It is the ability to reproduce a consistent result in time and space, or

from different observers

•Validity:

It refers to the fact that a tool measures exactly what it proposes to

measure

It’s a multiple testing correction method. Benjamini and Hochberg provide in a
simple stepwise procedure (BH) that controls the FDR when the test statistics are
statistically independent.

If the p value is set to 0.05, there is a 5% error margin for each single gene to pass
the test.

If 100 genes are tested, 5 genes could be found to be significant by chance, even
though they are not.

If testing a group of 10,000 genes, 500 might be found to be significant by
chance.

Therefore, it is important to correct the p-value of each gene when performing a
statistical test on a group or genes.

Multiple testing correction adjusts the individual p-value for each gene to keep the
overall error rate to less than or equal to the user-specified p-cutoff value.

0.05*i/m [i=rank, m= total no of variables]

FDR, FWE

Multiple testing refers to conducting more than one hypothesis test simultaneously
to make inferences on different aspects of a problem in a study experiment.

The significance level of a P value is defined under a single test.

When more than one test is conducted, simple use of the significance level for each
individual test leads to a probability of false-positive findings that is greater than
the stated α level.

Two commonly used approaches for choosing a significance level in multiple
testing are the FEW (“family-wise error rate”) and the FDR (False discovery rate).

False discovery rate
When a study involves a large number of tests, the FDR error measure is a more
useful approach in determining a significance cutoff because the FWE approach is
too stringent.

False discovery rate (FDR) is the expected proportion of the null hypotheses that
are falsely rejected divided by the total number of rejections.

Expected proportion of type 1 errors.