Monroe L. Weber-Shirk S chool of Civil and Environmental Engineering “BAD” DATA Sun e e e e

Monroe L. Weber-Shirk

School of Civil and

Environmental Engineering

“BAD” DATA“BAD” DATA

SunSun eeee SunSun

OverviewOverview

Bad Data Learning from unexpected results Isotherm Analysis Research

Bad Data Learning from unexpected results Isotherm Analysis Research

Sources of “Bad” DataSources of “Bad” Data Error in preparation of samples

mass or volume measurement error contamination improper storage sample substitution sample loss samples with high heterogeneity

Apparatus failures leaks incompatible materials inadequate control of an important parameter

Error in preparation of samples mass or volume measurement error contamination improper storage sample substitution sample loss samples with high heterogeneity

Apparatus failures leaks incompatible materials inadequate control of an important parameter

Instrument ErrorsInstrument Errors

detector malfunction below detection limit or above maximum interference software (instrument or computer) hardware (analog to digital converter, power

supply,...) calibration

detector malfunction below detection limit or above maximum interference software (instrument or computer) hardware (analog to digital converter, power

supply,...) calibration

More sources of “Bad Data”More sources of “Bad Data”

Error in data analysis numerical error (data entry) units (classic errors of factors of 10 and factors of

1000) incorrectly applied theory

Error in theory

Error in data analysis numerical error (data entry) units (classic errors of factors of 10 and factors of

1000) incorrectly applied theory

Error in theory

Bad Data aren’t Bad!

“Bad” data usually means the results were unexpected perhaps unorthodox!

Copernicus “Concerning the Revolutions of the Celestial Bodies”1543

Papal Index of forbidden books until 1835 _____________________

Data do not lie! Data always mean something If you ignore data that you don’t understand you are

missing an opportunity to learn

Bad data for 292 years!

Unexpected Results

Lack of repeatability (poor precision) scatter for all data outlier systemic error

0123456789

10

0 2 4 6 8 10

distance from source (m)

Con

cen

trati

on

(m

g/L

)

measured concentration

expected concentration

0123456789

10

0 2 4 6 8 10

distance from source (m)

Con

cen

trat

ion

(m

g/L

)

measured concentration

expected concentration

0123456789

10

0 2 4 6 8 10

distance from source (m)

Con

cen

trat

ion

(m

g/L

) measured concentration

expected concentration

Unexpected Results

Inconsistent with theory mass balances indicate loss or gain of mass

inconsistent with previous results

Sun ee Sun

02468

101214161820

0 2 4 6 8 10

parameter A

Par

amet

er B

measured trend

expected trend

Responses to Unexpected Results

Determine accuracy of technique by analyzing known samples

Determine precision of technique by analyzing replicates

Evaluate propagation of errors through analysis are you trying to measure the difference between two large

numbers? is the precision of the measurement similar to the magnitude

of the estimate? Are you not controlling an important parameter? Is the parameter that you are studying insignificant?

Isotherm Analysis Pointers Units

Express mass of VOC in grams Express concentrations as g per mL Remember GC injection volume was 0.1 mL

Use names to keep track of parameters in spreadsheet

Build sheet from left to right

More PointersMore Pointers

Soil density = 1.6 g/mL Soil moisture content is 13% Soil mass was close to 20 g Analyze the 4 data sets as sets

Use the data from one group to calculate a single value for each parameter

You will get 4 estimates for each parameter

Soil density = 1.6 g/mL Soil moisture content is 13% Soil mass was close to 20 g Analyze the 4 data sets as sets

Use the data from one group to calculate a single value for each parameter

You will get 4 estimates for each parameter

L

GL

GGGVOC

GS

GLS

L VH

CVCM

CM

HK

sc

spreadsheet

Why is a small fs a measurement problem?

Why is a small fs a measurement problem?

EPICS Error AnalysisEPICS Error Analysis

water water

low solubilit

y

low solubilit

y

high solubilit

y

high solubilit

y

Assume 10% error in measuring gas concentrations

What are the maximum and minimum values of mass in liquid phase?

Assume 10% error in measuring gas concentrations

What are the maximum and minimum values of mass in liquid phase?

(10 ± 1) – (9 ± 0.9)

11 – 8.1 = 2.9

9 – 9.9 = -0.9

(10 ± 1) –( 1 ± 0.1)

11 – 0.9 = 10.1

9 – 1.1 = 7.9