| |
| |
Preface | |
| |
| |
| |
Fitting data with nonlinear regression | |
| |
| |
| |
An example of nonlinear regression | |
| |
| |
Example data | |
| |
| |
| |
Clarify your goal. Is nonlinear regression the appropriate analysis? | |
| |
| |
| |
Prepare your data and enter it into the program | |
| |
| |
| |
Choose your model | |
| |
| |
| |
Decide which model parameters to fit and which to constrain | |
| |
| |
| |
Choose a weighting scheme | |
| |
| |
| |
Choose initial values | |
| |
| |
| |
Perform the curve fit and interpret the best-fit parameter values | |
| |
| |
| |
Preparing data for nonlinear regression | |
| |
| |
Avoid Scatchard, Lineweaver-Burk, and similar transforms whose goal is to create a straight line | |
| |
| |
Transforming X values | |
| |
| |
Don't smooth your data | |
| |
| |
Transforming Y values | |
| |
| |
Change units to avoid tiny or huge values | |
| |
| |
Normalizing | |
| |
| |
Averaging replicates | |
| |
| |
Consider removing outliers | |
| |
| |
| |
Nonlinear regression choices | |
| |
| |
Choose a model for how Y varies with X | |
| |
| |
Fix parameters to a constant value? | |
| |
| |
Initial values | |
| |
| |
Weighting | |
| |
| |
Other choices | |
| |
| |
| |
The first five questions to ask about nonlinear regression results | |
| |
| |
Does the curve go near your data? | |
| |
| |
Are the best-fit parameter values plausible? | |
| |
| |
How precise are the best-fit parameter values? | |
| |
| |
Would another model be more appropriate? | |
| |
| |
Have you violated any of the assumptions of nonlinear regression? | |
| |
| |
| |
The results of nonlinear regression | |
| |
| |
Confidence and prediction bands | |
| |
| |
Correlation matrix | |
| |
| |
Sum-of-squares | |
| |
| |
R[superscript 2] (coefficient of determination) | |
| |
| |
Does the curve systematically deviate from the data? | |
| |
| |
Could the fit be a local minimum? | |
| |
| |
| |
Troubleshooting "bad" fits | |
| |
| |
Poorly defined parameters | |
| |
| |
Model too complicated | |
| |
| |
The model is ambiguous unless you share a parameter | |
| |
| |
Bad initial values | |
| |
| |
Redundant parameters | |
| |
| |
Tips for troubleshooting nonlinear regression | |
| |
| |
| |
Fitting data with linear regression | |
| |
| |
| |
Choosing linear regression | |
| |
| |
The linear regression model | |
| |
| |
Don't choose linear regression when you really want to compute a correlation coefficient | |
| |
| |
Analysis choices in linear regression | |
| |
| |
X and Y are not interchangeable in linear regression | |
| |
| |
Regression with equal error in X and Y | |
| |
| |
Regression with unequal error in X and Y | |
| |
| |
| |
Interpreting the results of linear regression | |
| |
| |
What is the best-fit line? | |
| |
| |
How good is the fit? | |
| |
| |
Is the slope significantly different from zero? | |
| |
| |
Is the relationship really linear? | |
| |
| |
Comparing slopes and intercepts | |
| |
| |
How to think about the results of linear regression | |
| |
| |
Checklist: Is linear regression the right analysis for these data? | |
| |
| |
| |
Models | |
| |
| |
| |
Introducing models | |
| |
| |
What is a model? | |
| |
| |
Terminology | |
| |
| |
Examples of simple models | |
| |
| |
| |
Tips on choosing a model | |
| |
| |
Overview | |
| |
| |
Don't choose a linear model just because linear regression seems simpler than nonlinear regression | |
| |
| |
Don't go out of your way to choose a polynomial model | |
| |
| |
Consider global models | |
| |
| |
Graph a model to understand its parameters | |
| |
| |
Don't hesitate to adapt a standard model to fit your needs | |
| |
| |
Be cautious about letting a computer pick a model for you | |
| |
| |
Choose which parameters, if any, should be constrained to a constant value | |
| |
| |
| |
Global models | |
| |
| |
What are global models? | |
| |
| |
| |
Fitting incomplete data sets | |
| |
| |
| |
The parameters you care about cannot be determined from one data set | |
| |
| |
Assumptions of global models | |
| |
| |
How to specify a global model | |
| |
| |
| |
Compartmental models and defining a model with a differential equation | |
| |
| |
What is a compartmental model? What is a differential equation? | |
| |
| |
Integrating a differential equation | |
| |
| |
The idea of numerical integration | |
| |
| |
More complicated compartmental models | |
| |
| |
| |
How nonlinear regression works | |
| |
| |
| |
Modeling experimental error | |
| |
| |
Why the distribution of experimental error matters when fitting curves | |
| |
| |
Origin of the Gaussian distribution | |
| |
| |
From Gaussian distributions to minimizing sums-of-squares | |
| |
| |
Regression based on nongaussian scatter | |
| |
| |
| |
Unequal weighting of data points | |
| |
| |
Standard weighting | |
| |
| |
Relative weighting (weighting by 1/Y[superscript 2]) | |
| |
| |
Poisson weighting (weighting by 1/Y) | |
| |
| |
Weighting by observed variability | |
| |
| |
Error in both X and Y | |
| |
| |
Weighting for unequal number of replicates | |
| |
| |
Giving outliers less weight | |
| |
| |
| |
How nonlinear regression minimizes the sum-of-squares | |
| |
| |
Nonlinear regression requires an iterative approach | |
| |
| |
How the nonlinear regression method works | |
| |
| |
Independent scatter | |
| |
| |
| |
Confidence intervals of the parameters | |
| |
| |
| |
Asymptotic standard errors and confidence intervals | |
| |
| |
Interpreting standard errors and confidence intervals | |
| |
| |
How asymptotic standard errors are computed | |
| |
| |
An example | |
| |
| |
Because asymptotic confidence intervals are always symmetrical, it matters how you express your model | |
| |
| |
Problems with asymptotic standard errors and confidence intervals | |
| |
| |
What if your program reports "standard deviations" instead of "standard errors"? | |
| |
| |
How to compute confidence intervals from standard errors | |
| |
| |
| |
Generating confidence intervals by Monte Carlo simulations | |
| |
| |
An overview of confidence intervals via Monte Carlo simulations | |
| |
| |
Monte Carlo confidence intervals | |
| |
| |
Perspective on Monte Carlo methods | |
| |
| |
How to perform Monte Carlo simulations with Prism | |
| |
| |
Variations of the Monte Carlo method | |
| |
| |
| |
Generating confidence intervals via model comparison | |
| |
| |
Overview on using model comparison to generate confidence intervals | |
| |
| |
A simple example with one parameter | |
| |
| |
Confidence interval for the sample data with two parameters | |
| |
| |
Using model comparison to generate a confidence contour for the example data | |
| |
| |
Converting the confidence contour into confidence intervals for the parameters | |
| |
| |
How to use Excel's solver to adjust the value of a parameter to get the desired sum-of-squares | |
| |
| |
More than two parameters | |
| |
| |
| |
Comparing the three methods for creating confidence intervals | |
| |
| |
Comparing the three methods for our first example | |
| |
| |
A second example. Enzyme kinetics | |
| |
| |
A third example | |
| |
| |
Conclusions | |
| |
| |
| |
Using simulations to understand confidence intervals and plan experiments | |
| |
| |
| |
Should we express the middle of a dose-response curve as EC[subscript 50] or log(EC[subscript 50])? | |
| |
| |
| |
Exponential decay | |
| |
| |
How to generate a parameter distribution with Prism | |
| |
| |
| |
Comparing models | |
| |
| |
| |
Approach to comparing models | |
| |
| |
Why compare models? | |
| |
| |
Before you use a statistical approach to comparing models | |
| |
| |
Statistical approaches to comparing models | |
| |
| |
| |
Comparing models using the extra sum-of-squares F test | |
| |
| |
Introducing the extra sum-of-squares F test | |
| |
| |
The F test is for comparing nested models only | |
| |
| |
How the extra sum-of-squares F test works | |
| |
| |
How to determine a P value from F | |
| |
| |
| |
Comparing models using Akaike's Information Criterion (AIC) | |
| |
| |
Introducing Akaike's Information Criterion (AIC) | |
| |
| |
How AIC compares models | |
| |
| |
A second-order (corrected) AIC | |
| |
| |
The change in AICc tells you the likelihood that a model is correct | |
| |
| |
The relative likelihood or evidence ratio | |
| |
| |
Terminology to avoid when using AIC[subscript c] | |
| |
| |
How to compare models with AICc by hand | |
| |
| |
One-way ANOVA by AICc | |
| |
| |
| |
How should you compare models--AIC[subscript c] or F test? | |
| |
| |
A review of the approaches to comparing models | |
| |
| |
Pros and cons of using the F test to compare models | |
| |
| |
Pros and cons of using AIC[subscript c] to compare models | |
| |
| |
Which method should you use? | |
| |
| |
| |
Examples of comparing the fit of two models to one data set | |
| |
| |
| |
Two-site competitive binding model clearly better | |
| |
| |
| |
Two-site binding model doesn't fit better | |
| |
| |
| |
Can't get a two-site binding model to fit at all | |
| |
| |
| |
Testing whether a parameter differs from a hypothetical value | |
| |
| |
Example. Is the Hill slope factor statistically different from 1.0? | |
| |
| |
Compare models with the F test | |
| |
| |
Compare models with AIC[subscript c] | |
| |
| |
Compare with t test | |
| |
| |
| |
How does a treatment change the curve? | |
| |
| |
| |
Using global fitting to test a treatment effect in one experiment | |
| |
| |
Does a treatment change the EC[subscript 50]? | |
| |
| |
Does a treatment change the dose-response curve? | |
| |
| |
| |
Using two-way ANOVA to compare curves | |
| |
| |
Situations where curve fitting isn't helpful | |
| |
| |
Introduction to two-way ANOVA | |
| |
| |
How ANOVA can compare "curves" | |
| |
| |
Post-tests following two-way ANOVA | |
| |
| |
The problem with using two-way ANOVA to compare curves | |
| |
| |
| |
Using a paired t test to test for a treatment effect in a series of matched experiments | |
| |
| |
The advantage of pooling data from several experiments | |
| |
| |
An example. Does a treatment change logEC[subscript 50]? Pooling data from three experiments | |
| |
| |
Comparing via paired t test | |
| |
| |
Why the paired t test results don't agree with the individual comparisons | |
| |
| |
| |
Using global fitting to test for a treatment effect in a series of matched experiments | |
| |
| |
Why global fitting? | |
| |
| |
Setting up the global model | |
| |
| |
Fitting the model to our sample data | |
| |
| |
Was the treatment effective? Fitting the null hypothesis model | |
| |
| |
| |
Using an unpaired t test to test for a treatment effect in a series of unmatched experiments | |
| |
| |
An example | |
| |
| |
Using the unpaired t test to compare best-fit values of V[subscript max] | |
| |
| |
| |
Using global fitting to test for a treatment effect in a series of unmatched experiments | |
| |
| |
Setting up a global fitting to analyze unpaired experiments | |
| |
| |
Fitting our sample data to the global model | |
| |
| |
Comparing models with an F test | |
| |
| |
Comparing models with AIC[subscript c] | |
| |
| |
Reality check | |
| |
| |
| |
Fitting radioligand and enzyme kinetics data | |
| |
| |
| |
The law of mass action | |
| |
| |
What is the law of mass action? | |
| |
| |
The law of mass action applied to receptor binding | |
| |
| |
Mass action model at equilibrium | |
| |
| |
Fractional occupancy predicted by the law of mass action at equilibrium | |
| |
| |
Assumptions of the law of mass action | |
| |
| |
Hyperbolas, isotherms, and sigmoidal curves | |
| |
| |
| |
Analyzing radioligand binding data | |
| |
| |
Introduction to radioligand binding | |
| |
| |
Nonspecific binding | |
| |
| |
Ligand depletion | |
| |
| |
| |
Calculations with radioactivity | |
| |
| |
Efficiency of detecting radioactivity | |
| |
| |
Specific radioactivity | |
| |
| |
Calculating the concentration of the radioligand | |
| |
| |
Radioactive decay | |
| |
| |
Counting errors and the Poisson distribution | |
| |
| |
The GraphPad radioactivity web calculator | |
| |
| |
| |
Analyzing saturation radioligand binding data | |
| |
| |
Introduction to saturation binding experiments | |
| |
| |
Fitting saturation binding data | |
| |
| |
Checklist for saturation binding | |
| |
| |
Scatchard plots | |
| |
| |
Analyzing saturation binding with ligand depletion | |
| |
| |
| |
Analyzing competitive binding data | |
| |
| |
What is a competitive binding curve? | |
| |
| |
Competitive binding data with one class of receptors | |
| |
| |
Shallow competitive binding curves | |
| |
| |
Competitive binding with two receptor types (different K[subscript d] for hot ligand) | |
| |
| |
Heterologous competitive binding with ligand depletion | |
| |
| |
| |
Homologous competitive binding curves | |
| |
| |
Introducing homologous competition | |
| |
| |
Theory of homologous competition binding | |
| |
| |
Why homologous binding data can be ambiguous | |
| |
| |
Using global curve fitting to analyze homologous (one site) competition data | |
| |
| |
Analyzing homologous (one site) competition data without global curve fitting | |
| |
| |
Homologous competitive binding with ligand depletion | |
| |
| |
Fitting homologous competition data (two sites) | |
| |
| |
| |
Analyzing kinetic binding data | |
| |
| |
Dissociation ("off rate") experiments | |
| |
| |
Association binding experiments | |
| |
| |
Fitting a family of association kinetic curves | |
| |
| |
Globally fitting an association curve together with a dissociation curve | |
| |
| |
Analysis checklist for kinetic binding experiments | |
| |
| |
Using kinetic data to test the law of mass action | |
| |
| |
Kinetics of competitive binding | |
| |
| |
| |
Analyzing enzyme kinetic data | |
| |
| |
Introduction to enzyme kinetics | |
| |
| |
How to determine V[subscript max] and K[subscript m] | |
| |
| |
Comparison of enzyme kinetics with radioligand binding | |
| |
| |
Displaying enzyme kinetic data on a Lineweaver- Burk plot | |
| |
| |
Allosteric enzymes | |
| |
| |
Enzyme kinetics in the presence of an inhibitor | |
| |
| |
| |
Fitting dose-response curves | |
| |
| |
| |
Introduction to dose-response curves | |
| |
| |
What is a dose-response curve? | |
| |
| |
The equation for a dose-response curve | |
| |
| |
Other measures of potency | |
| |
| |
Dose-response curves where X is concentration, not log of concentration | |
| |
| |
Why you should fit the logEC[subscript 50] rather than EC[subscript 50] | |
| |
| |
Decisions when fitting sigmoid dose-response curves | |
| |
| |
Checklist: Interpreting a dose-response curve | |
| |
| |
| |
The operational model of agonist action | |
| |
| |
Limitations of dose-response curves | |
| |
| |
Derivation of the operational model | |
| |
| |
Shallower and steeper dose-response curves | |
| |
| |
Designing experiments to fit to the operational model | |
| |
| |
Fitting the operational model to find the affinity and efficacy of a full agonist | |
| |
| |
Fitting the operational model to find the affinity and efficacy of a partial agonist | |
| |
| |
| |
Dose-response curves in the presence of antagonists | |
| |
| |
Competitive antagonists | |
| |
| |
Using global fitting to fit a family of dose-response curves to the competitive interaction model | |
| |
| |
Fitting agonist EC[subscript 50] values to the competitive interaction model | |
| |
| |
Antagonist inhibition curves | |
| |
| |
| |
Complex dose-response curves | |
| |
| |
Asymmetric dose-response curves | |
| |
| |
Bell-shaped dose-response curves | |
| |
| |
Biphasic dose-response curves | |
| |
| |
| |
Fitting curves with GraphPad Prism | |
| |
| |
| |
Nonlinear regression with Prism | |
| |
| |
Using Prism to fit a curve | |
| |
| |
Which choices are most fundamental when fitting curves? | |
| |
| |
Prism's nonlinear regression error messages | |
| |
| |
| |
Constraining and sharing parameters | |
| |
| |
The constraints tab of the nonlinear regression parameters dialog | |
| |
| |
Constraining to a constant value | |
| |
| |
Data set constants | |
| |
| |
Constrain to a range of values | |
| |
| |
Shared parameters (global fitting) | |
| |
| |
| |
Prism's nonlinear regression dialog | |
| |
| |
The equation tab | |
| |
| |
Comparison tab | |
| |
| |
Initial values tab | |
| |
| |
Constraints for nonlinear regression | |
| |
| |
Weighting tab | |
| |
| |
Output tab | |
| |
| |
Range tab | |
| |
| |
Default preferences for nonlinear regression | |
| |
| |
| |
Classic nonlinear models built into Prism | |
| |
| |
Equilibrium binding | |
| |
| |
Dose-response | |
| |
| |
Exponential | |
| |
| |
Other classic equations | |
| |
| |
| |
Importing equations and equation libraries | |
| |
| |
Selecting from the equation library | |
| |
| |
Adding equations to the equation library | |
| |
| |
Importing equations | |
| |
| |
| |
Writing user-defined models in Prism | |
| |
| |
What kinds of equations can you enter? | |
| |
| |
Equation syntax | |
| |
| |
Available functions for user-defined equations | |
| |
| |
Using the IF function | |
| |
| |
How to fit different portions of the data to different equations | |
| |
| |
How to define different models for different data sets | |
| |
| |
Defining rules for initial values and constraints | |
| |
| |
Managing your list of equations | |
| |
| |
Modifying equations | |
| |
| |
| |
Linear regression with Prism | |
| |
| |
Entering data for linear regression | |
| |
| |
Choosing a linear regression analysis | |
| |
| |
Default preferences for linear regression | |
| |
| |
Using nonlinear regression to fit linear data | |
| |
| |
Deming (Model II) linear regression | |
| |
| |
Inverse linear regression with Prism | |
| |
| |
| |
Reading unknowns from standard curves | |
| |
| |
Introduction to standard curves | |
| |
| |
Determining unknown concentrations from standard curves | |
| |
| |
Standard curves with replicate unknown values | |
| |
| |
Potential problems with standard curves | |
| |
| |
| |
Graphing a family of theoretical curves | |
| |
| |
Creating a family of theoretical curves | |
| |
| |
| |
Fitting curves without regression | |
| |
| |
Introducing spline and lowess | |
| |
| |
Spline and lowess with Prism | |
| |
| |
Annotated bibliography | |