Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237385797

Nonlinear and Linear Regression Applied to Concentration versus Time


Kinetic Data from Pinhas's Sanitizer Evaporation Project

Article  in  Journal of Chemical Education · November 2011


DOI: 10.1021/ed101040d

CITATIONS READS

11 954

1 author:

Todd P Silverstein
Willamette University
120 PUBLICATIONS   844 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

ADH kinetics lab View project

TiO2 and sunscreen View project

All content following this page was uploaded by Todd P Silverstein on 11 August 2019.

The user has requested enhancement of the downloaded file.


COMMUNICATION

pubs.acs.org/jchemeduc

Nonlinear and Linear Regression Applied to Concentration versus


Time Kinetic Data from Pinhas’s Sanitizer Evaporation Project
Todd P. Silverstein*
Chemistry Department, Willamette University, Salem, Oregon 97301, United States

ABSTRACT: A recent article in this Journal described a laboratory experiment in which students study the evaporation kinetics of
hand sanitizer. In this communication, the differences between using linear regression on linearized kinetics data versus nonlinear
regression on raw data are examined, and arguments in favor of the latter are presented.
KEYWORDS: First-Year Undergraduate/General, High School/Introductory Chemistry, Laboratory Instruction, Physical
Chemistry, Hands-On Learning/Manipulatives, Inquiry-Based/Discovery Learning, Kinetics, Phases/Phase Transitions/Diagrams,
Rate Law

R ecently in this Journal, Pinhas published a wonderful, simple


laboratory project studying the evaporation kinetics of hand
sanitizer.1 A few suggestions are given to supplement the data
Second, Pinhas plotted ln(mass0/masst) versus t and 1/
masst  1/mass0 versus t (Figures 2 and 3, respectively, in his
article). Because mass0 is equivalent to [reactant]0, and masst is
analysis of this experiment. First, writing the equation for the equivalent to [reactant]t, Figure 2 is essentially a standard semi-
evaporation phase-change reaction is instructive: log plot suitable for linearizing first-order kinetic data. Likewise,
Figure 3 is essentially a standard inverse plot suitable for
sanitizerðliq or gelÞ f sanitizerðgÞ linearizing second-order kinetic data. Students can easily use
these linearized plots to conclude that the reaction is first order.
This equation helps the students understand that the measured Furthermore, as explained in all introductory chemistry text-
mass of the condensed-phase sanitizer at different times can be books, the slope of the linear plot can be used to determine the
expressed with mass0 as the initial mass of reactant (liquid or gel rate constant, k, for the reaction. However, this is generally not
sanitizer) measured at time zero; masst as the mass of remaining the most accurate way to get k.
reactant at time t; and (mass0  masst) as the mass of product (i.e., Nonlinear fitting software (e.g., Excel Solver, Kaleidagraph,
vapor) at time t. From this perspective, it is apparent that when SigmaPlot) is often available on school and university computers.
Pinhas plotted (mass0  masst) versus time (Figure 1 in his In fitting experimental results, nonlinear regression is preferable
article), he plotted the increase in the product with time. The plot because it avoids skewing due to mathematical manipulation of the
is not linear, and thus, the reaction is not zeroth order. raw data ( refs 2, 3 and references cited therein). Furthermore, it is

Figure 1. Nonlinear regression on raw data (mass of sanitizer remaining at time t) for (A) gel and (B) liquid sanitizer. Fit equations are masst = mass0ekt for
first order and masst = [(1/mass0) + kt]1 for second order. Best fit values for the rate constants, k, are given in min1, and also converted to h1. Uncertainty
values (() are standard errors obtained from nonlinear (Kaleidagraph) and linear (Excel) regression.

Published: August 18, 2011

Copyright r 2011 American Chemical Society and


Division of Chemical Education, Inc. 1589 dx.doi.org/10.1021/ed101040d | J. Chem. Educ. 2011, 88, 1589–1590
Journal of Chemical Education COMMUNICATION

Table 1. Comparison of Linear versus Nonlinear Regression ’ REFERENCES


Component Linear Regressiona Nonlinear Regressionb (1) Pinhas, A. R. A Kinteic Study Using Evaporation of Different
Types of Hand-Rub Sanitizers. J. Chem. Educ. 2010, 87 (9), 950–951.
First-order plot axes Ln(mass0/masst) vs time Mass vs time (2) Silverstein, T. P. Using a Graphing Calculator to Determine a
Curve type Linear increase Exponential decay First-Order Rate Constant. J. Chem. Educ. 2004, 81 (4), 485.
(3) Silverstein, T. P. Quantitative Determination of DNA-Ligand
Equation fit Ln(mass0/masst) = kt Masst = mass0ekt
Binding: Improved Data Analysis by Nonlinear Regression. J. Chem.
k (h1) for gel 4.559 ( 0.012 4.523 ( 0.022 Educ. 2008, 85 (9), 1192–1193.
τ (min) for gel 9.122 ( 0.024 9.19 ( 0.04
R2 0.9996 0.9997
k (h1) for liquid 5.52 ( 0.05 5.57 ( 0.08
τ (min) for liquid 7.53 ( 0.07 7.46 ( 0.11
R2 0.9957 0.9975
a
Data are from Figure 2 in ref 1. b Raw data from Figure 2 in ref 1 are
used to calculate the parameters.

often simpler to apply nonlinear regression to the raw data (masst


versus t), as opposed to linearizing the data.
An analysis using nonlinear regression is applied to the gel and
liquid data sets that Pinhas published. The results from nonlinear
regression applied to the raw data for the gel and liquid sanitizers
for first-order (exponential) as well as second-order decline in
mass with time are shown in Figure 1. The decline of mass with
time demonstrates that the reaction is not zeroth order and the
first-order fits are clearly superior to second-order, as Pinhas also
showed with his linearized plots.1 The rate constant and half-life
for the vaporization of gel and liquid sanitizer, as determined by
semi-log linearized data and noninear regression, are listed in
Table 1. The main advantage of nonlinear regression applied to
raw data is that any skewing due to mathematical manipulation of
the data (taking logs, reciprocals, etc.) is avoided (refs 2, 3, and
references cited therein). A further minor advantage observed
here is that the nonlinear regression fits have slightly better R2
values. Although kliquid obtained from linear regression is less
than kliquid (nonlinear regression) by a statistically insignificant
amount (5.52 ( 0.05 vs 5.57 ( 0.08 h1), the same cannot be
said for kgel. The value from linear regression is higher by 3 the
uncertainty: 4.559 ( 0.012 versus 4.523 ( 0.022 h1. This
difference is statistically significanta and would require the
students to choose which value is the more reliable one. As
explained above, this is almost always the value determined from
nonlinear regression. It is interesting to note that this statistically
significant difference between results from linear versus non-
linear regression is found even when using good linearized data
sets such as those shown in Pinhas’s Figure 2. This strengthens
the argument for using nonlinear regression in data analysis.

’ AUTHOR INFORMATION
Corresponding Author
*E-mail: tsilvers@willamette.edu.

’ ADDITIONAL NOTE
a
Statistical significance here applies only to results obtained from
this one data set. In a research setting, students would obtain
replicate results and average them. In this case, the standard
deviation would undoubtedly be higher than the standard errors
found here. Nevertheless, the fact that nonlinear and linear
regression applied to a single data set give statistically signifi-
cantly different results is instructive.
1590 dx.doi.org/10.1021/ed101040d |J. Chem. Educ. 2011, 88, 1589–1590

View publication stats

You might also like