Lecture Exam 2 Flashcards
92
Sensitivity
Testing for true positives
Specificity
Testing for true negatives
T test graph
A graph that shows how data from each group differ in averages and using that to understand the results of the t test
F test
used to compare the variances of two or more groups or samples
X axis
represents the variable being measured or studied.
y axis
represents the frequency or density of observations for each value or category on the x-axis
Group Means
The mean of each group is represented by a line or a bar on the graph
Error bars
represent standard errors, standard deviations, or confidence intervals.
T value
This value quantifies the difference between the means of the two groups relative to the variability within the groups
Critical value or significance level
This value quantifies the difference between the means of the two groups relative to the variability within the groups
Distribution curves in t test
The curves with more degrees of freedom are taller and have thinner tails
F value
This value quantifies the variability between group means relative to the variability within groups.
Guissan Distribution
When the mean, median, and mode are all the same values to make a normal distribution
Precision
occurs when a consistent answer comes from repeated tests; consider a target with shots scattered all around, but the average position of all shots is near the bullseye.
Accuracy
Imagine a target with all shots clustered tightly but far from the bullseye. This scenario represents high accuracy
Reliability
A result that has both accuracy and precision
Variance
The precision of a group of numbers - How precise?
Variance formula
Each individual data point in the sample is subtracted by the mean and then squared. Add all values together and divide it by n - 1
Standard Deviation
Most frequent used measure of precision and is found by square rooting the variance
Standard Deviation Ranges
+-1SD range: Includes 68% of the data points, extending one standard deviation above and below the mean.
+-2SD range: This would include about 95% of the data points, extending two standard deviations above and below the mean.
+-3SD range: This would cover about 99.7% of the data points, extending three standard deviations above and below the mean.
Coefficient of Variation formula
s/mean x 100% = C.V
Coefficient of Variation
Used to compare to sets, and the lower the % of variable distribution the more precise
Quality assurance
A program/process that assures accurate results, monitors methods and instruments and proposes troubleshooting methods
3 activities of quality assurance
pre-analytical (wrong patient info), analytical (off measured) , post-analytical (misinformed reporting)
Quality control Material
material that is of the same matrix of patient sample and analyzed with sample ( use at least 2 different “levels” like normal and abnormal)
Quality control
Actual tests that give numerical statistical measures of error
Quality control analysis
Using a Levey-Jennings chart to record QC results and determine if method is control or not using Westgards multirules
What kind of errors can be predicted by a bell curve chart?
Random errors
Patient Test Management
Monitor sample from start to finish
CLIA ‘88
Mandates 2 levels of every assay type, and is run once per day or once per shift; the results are analyzed on LJ chart
What does a shift look like on a LJ chart?
When all QC values are all distributed on one side of the mean for 5-7 consecutive days
What kind of errors are shifts and trends caused by?
Systematic errors
Lyophilized
freeze dried and reconsituted with water
Prediluted
ready to use
Ethylene-glycol based
liquid at 0 degrees Celsius
What does a trend look like on a LJ chart?
Occurs when QC results either decrease or increase consistently over 5-7 days
Westgard Multirules definition
set of quality control rules to interpret quality control results in the clinical laboratory
Westgard Multirules
1_2s, 1_3s, 2 of 3_2s, R_4s, 4_1s, 9/10x
Random error vs systematic error
Random error is from lack of precision and systematic error relates to lack of accuracy
Which Westgard rule is usually violated because of a random error?
R4s
How to get good test results?
Calibration using primary (reference materials) and secondary standards (calibrated from primary)
“Reference” methods
these are the gold standards for extremely accurate and reliable
Sensitivity formula
TP/TP+FN x 100
Specificity formula
TN/TN+FP x 100
Which federal law mandates specificity and sensitivity tests?
FDA, CLIA, CMS
FDA
Food and Drug Administration
CLIA
Clinical Laboratory Improvement Amendments
CMS
Center for Medicare and Medicaid Services
Examples of reference methods
microhematocrit, single channel cell counter, spectrophotometer
“Definitive” methods
using sophisticated instruments that measure a specific substance of interest in its pure form
A calibrator has _____ values
assigned
Control Runs
used with QC material that has the same matrix as bio sample
Commercial Controls
Not used as calibrators but used to monitor all procedures using 2 to 3 levels (normal, low abnormal, high abnormal) with assayed values and SD ranges
Assayed values
the established values of a particular substance
Efficiency
TP + TN / (TP + FP + TN + FN) x 100
Predictive value
Positive Predictive value: TP/TP+FP x 100
Negative Predictive value: TN/TN+FN x 100
Name 3 QC Materials
Lyophilized plasma, whole blood QC materials, body fluid control
Null Hypothesis (H0)
no statistical significant difference between glucose results from both analyzers
P= 0.05
Is the probability that the null h. is rejected when it is true
Alternate hypothesis (Ha)
statistically significant difference between glucose results from both analyzers
Two-tailed test
examine whether there is a difference between groups or conditions in any direction by using H0 and Ha (H1)
Paired T test
compares related measurements to assess whether there is a difference between paired data points
Unpaired T test
There is no direct link or matching between data points in one group and data points in the other group; can be 1 tailed or 2 tailed
Linear Regression Analysis by the method of Least Squares
Results from two tested methods plotted on linear graph where the results from the reference method (x axis) and the test method (y axis); if both tests are identical, there is a perfect correlation(r) and a linear line
Linear Regression formula
yc = mx +b
yc = predicted y intercept derived
Precision study
mandated by the CLIA to be a part of the method validation process and can be determined by SD and C.V
F test
used when there are more than 2 means to compare
Spearman, Kendall, and Pearson
names associated with different methods of measuring the strength and direction of relationships between variables
Pearson correlation coefficient
often denoted as r, measures the linear relationship between two continuous variables. It ranges from -1 to +1, where:
- r=+1 indicates a perfect positive linear relationship,
- r=−1 indicates a perfect negative linear relationship,
- r=0 indicates no linear relationship.
Spearman’s (ρ (rho)) rank correlation coefficient and Kendall’s (tau (τ)) correlation coefficient
measures the strength and direction of the monotonic relationship between two variables (spearman is more commonly used)
Monotonic
refers to a specific pattern or direction of change in a relationship between variables
Critical values
z values in rejection area of a tail test
how to find z score based on confidence level?
1-critical value (confidence level divided by 100) / 2 and then look for number on z-score table, add it up and theres the z score
Formula for Least Squares method
y = mx + b
m = n (summation of x values) - summation of xy / n (summation of x^2) - (summation of x)^2
b = summation of y - m (summation of x) / n