You are on page 1of 1

Confidence Intervals in parameter estimation. Example.

Consider the speed of ants as a function of temperature. We want to see if this can be
modelled using Arrhenius equation

Data
T °C Speed (cm/s) T (K) 1/T ln(speed)
10 0.5 283 0.003534 -0.69
20 2 293 0.003413 0.69
30 3.4 303 0.0033 1.22
38 6.5 311 0.003215 1.87

Let us take (a) first two data points, (b) first three data points and (c) all four data points.
E1
When we fit the data to ln( speed )  ln(k0 )    , written as Y=intercept + slope  X 
R T 
we get intercept and slope. This can be obtained using software. e.g. Excel or Matlab.

In Matlab , use mdl = fitlm(x,y). We can also get the upper and lower bounds (95%
confidence interval) using coefCI(mdl) in Matlab. You can get same information in Excel, if
you use Data Analysis pack

(a) With two points we get

Parameter Value Lower Bound Upper bound


Intercept 39.744 -∞ +∞
Slope -11442.8 -∞ +∞
R-square 1

(b) With three points, we get


Parameter Value Lower Bound Upper bound
Intercept 28.508 -56 1113
Slope -8227.29 -32902 16447
R-square 0.947

(c) With all the four points, we get

Parameter Value Lower Bound Upper bound


Intercept 26.84 12 41
Slope -7745.74 -12050 -3441
R-square 0.967

You might also like