Method Eval

to summarize the important features of a group of data

measures of center, spread, and shape

mean (average)
median (middle pt)
mode (most frequent)

how far the data is distributed

distribution

mean, median, and mode are identical;
distribution is symmetrical (bell curve)

A bell shaped distribution of results from a normal sample population.
Theoretical distribution with finte mean and variance.

A. if number of events is large, describes physical events
B. shown under general assumptions that the distribution of random errors of observation takes on a normal distribution as number of obervations become large

ease of interpretibility

68% of results fall within +1 or -1
95% of results fall within +2 or -2
5% falls outisde of + / - 2

1/ 6 on each side (total 2/ 3)

1/ 40 (total 1/ 20)

1/ 200 (total 1/ 100)

68.3%
1 in 3

90%
1 in 10

95%
1 in 20

99%
1 in 100

mode
mean
median

range
variance
st dev
CV

by statistics that summarize location and dispersion;
MEAN is a measure of centrality of a set of data

value that occurs most often;
mode is the most frequent observation

value for which half of the remaining are above and half are below

Xmax - Xmin

whatever it is SQUARED

by using standard deviation (sq root of variance)

it goes down; st dev and N is indirectly proportional

mens height AND women's height in one graph (each one is gaussian)

Cannot compare 2 kinds of data if the units are different

if we want S to go DOWN 2 (improve) then N needs to go UP 4

coefficient of variation (RSD)

5:100 (1:20)

1:1600
(1/40 x 1/40=1/1600)

the degree of closeness of the measurements with each other;
getting consistnet results of repeated measurements

repeatability and reproducbility error

degree of closeness of the measurements to the target or reference value;
measuring near the target or true or ref value

bias error

the smaller the % the better your reproducibility and hence performance

a range
mean ± 2s

accurate

ability to reproduce SOMEONE ELSE'S results!

random error

increase # of measurements

no direction, unpredictable

systematic errors:
constant error
proportional error

gives est range of values which is likely to include known population parameter,
est range being calculated from a given set of sample data
(mean ±2s)

as 95%,
but can produce 90%, 99%, 99.9%

values that are consistently outside the statistical variation

33% samples reassayed
to see which values outside the limit due to statistical variation and which for other reasons

5% samples reassayed;
this is called 95% confidence limit

proportional
constant

present in all measurements
either positive or negative

instrument
reagent
operator techniques
environmental conditions

instrument calibration
reagent
method

always consistent and predictable
has direction

trend;
increased concentration, increased source of error etc

to ensure data collected remains valid overtime

systematic monitoring of analytic processes to detect analytic errors that occur during analysis

reporting of incorrect pt test results

by assaying stable control materials and comparing determined vs expected values

observed values of a control material over time in context of upper / lower limit controls

method was performed adequately

problems are developing

as means ± st dev

variation in technique

from factors that contribute to constant diff between measurements (pos or neg)

stable
range of concentration
has the same matrix of specimens

2 levels of control

3 levels of control

graphic representation to help monitor how a process changes over time

± 2s

one control > 2S
rejection warning

reject a run when ONE control observation > 3S

reject run when 2 consecutive control observations are on the same side of mean and > 2S

Reject run when 4 consecutive control obs are on the same side of mean and > 1S

Reject run if range or diff between the max and min control observations >4S

1 2S

poor precision has led to random error (RE)

deterioration of the reagent