B. Concepts, Principles, Application of Quality Assurance in Analytical Processes and Results

Question 1

When to use mean or median in determining the central value of replicates?

Answer 1

Mean - no outlier, outlier will affect the mean

Median - with outlier; outlier will not affect the median

Question 2

Precision vs accuracy

Answer 2

Precision - closeness of data taken with the same procedure

Accuracy - closeness of data to the true value

Question 3

Terms to describe precision:

Answer 3

1. standard deviation
2. variance
3. coefficient of variation

Question 4

Terms to describe accuracy:

Answer 4

1. absolute error - difference between measured value and true value (can be positive or negative)
2. relative error - [absolute error/true value] x100

Question 5

Three types of error

Answer 5

1. Random (indeterminate) errors - affect precision
2. Systematic (determinate) errors - affect accuracy
3. Gross error - often large, caused by human errors

Question 6

Define outliers and bias

Answer 6

Due to gross errors; data that differs significantly with other measurements

Bias - measures the systematic error
associated with an analysis. It has a
negative sign if it causes the results to
be low and a positive sign otherwise.

Question 7

Types of systematic errors

Answer 7

1. Instrumental errors - easily detectable
2. Method errors - most difficult to detect
3. Personal errors - due to personal judgement (prejudice or bias)

Question 8

The effect of systematic errors may be either ____ or _____

Answer 8

Constant error - absolute error is independent of the sample size but relative error varies

Proportional errors - proportional to the size of sample, but relative error stays constant

Question 9

The effect of a constant error becomes more serious as the size of the quantity measured _____.

One way of reducing the effect of constant error is to _____

Answer 9

decreases

increase the sample size until the error is acceptable.

Question 10

A common cause of proportional errors is the presence of _______ in the sample.

Answer 10

interfering contaminants

Question 11

How to correct instrument and personal errors?

Answer 11

1. Instrument error - mostly calibration and standardization
2. Personal errors - proper good laboratory practices, automated procedure

Question 12

How to correct instrument and personal errors?

Answer 12

1. Analyze standard reference materials (SRMs) from NIST
2. Independent Analysis - different independent and reliable analytical method
3. Blank determination - reveals errors due to contaminants
4. Variation in sample size

Question 13

How do you describe a normally distributed set of data?

Answer 13

Majority of the measurements are near the central (mean) value and form a bell shape curve with equal positive and negative distribution

Question 14

What is the standard error of the mean?

Answer 14

The standard deviation of a set of data divided by the square root of the number of data points in the set

how precise your sample mean is
↑SEM = less reliable

Question 15

Is it better to average more measurements to improve SEM or improve sample standard deviation?

Answer 15

improve sample standard deviation

Question 16

N must be _____ to consider as good estimator of population

Answer 16

N > 20

Question 17

What is variance, RSD and CV?

Answer 17

variance = sd^2

Both are relative term of standard deviation:
Relative Standard Deviation = sd/mean
RSD, ppt = sd/mean x 1000
Coefficient of Variation = RSD x 100

Question 18

Uncertainty propagation for addition and subtraction

Answer 18

error = √ [ea^2 + eb^2 + ec^2]

en = error/uncertainty of the measurement

Question 19

Uncertainty propagation for multiplication and division

Answer 19

error = value √ [(ea/a)^2 + (eb/b)^2 + …]

en = error/uncertainty of the measurement
n = measurement value

Question 20

Uncertainty propagation for a^x

Answer 20

error = value * x (ea/a)

x - exponent
en = error/uncertainty of the measurement
n = measurement value

Question 21

Uncertainty propagation for log a and antilog a

Answer 21

log a: error = (1/ln10) (ea/a)

antilog a: error = value ln10 ea

Question 22

Uncertainty propagation for e^x and ln a

Answer 22

e^x: error = value * ex
ln a: error = ea/a

Question 23

Retain the appropriate significant figure:
a. log 4.000 × 10^–5 = -4.3979400
b. antilog 12.5 = 3.162277 × 10^12

Answer 23

a. -4.3979 = 4 SF
b. 3 × 10^12 = 1 SF

Question 24

Type I vs Type II error

Answer 24

Type I (false positive) - Rejecting the null hypothesis when it is actually true
→ telling you’re positive in COVID even if you’re not

Type II (false negative) - Failing to reject the null hypothesis when it is actually false
→ telling you’re negative in COVID but you’re actually infected

Question 25

How to know if a data is an outlier?

Answer 25

Q-Test 1. Q = (outlier data - nearest data) / (upper limit - lower limit) 2. Count the number of observations 3. Look at the confidence level 4. Check Q table for Qcrit 5. Retain if Q < Qcrit

Question 26

In trace procedures, the reliability of results ______ dramatically with a decrease in analyte level.

Answer 26

usually decreases

Question 27

Define matrix effect

Answer 27

Interference cause by extraneous species in the matrix

Question 28

Steps in obtaining a laboratory sample.

Answer 28

1. Identify the population 2. Collect gross sample 3. Reduce to laboratory sample

Question 34

Goal of sampling process is to obtain a ____ sample that is representative of the population

Answer 34

random sample - equal chance of getting selected

Question 34

How to compute for the overall standard deviation?

Answer 34

So^2 = Ss^2 + Sm^2 Ss - SD due to sampling Sm - SD due to method

Question 34

If the sampling uncertainty is large and cannot be improved (____), it is often a good idea to _____ so that more samples can be analyzed in a given length of time.

Answer 34

Sm ≤ Ss /3 switch to a less precise but faster method of analysis More samples = improve precision

Question 35

How to obtain the number of particles needed to achieve a given relative standard deviation?

Answer 35

N = (1 - p) / (p RSD^2) p - probability of getting certain type of particle