You are on page 1of 5

BIOE 340: Modeling and Simulation in Bioengineering

Review for Midterm

1. A fitted model with more predictors will generally have a lower Training Set Error than a

model with fewer predictors. True or False.

Error

# Predictors

2. In a linear regression with several variables, if the F statistic is significant, all of the predictors

have statistically significant effects. True or False.

3. Test set error computed from 5-fold cross validation can be used to select best model. True

or False. (Training test deseydi yanlış)

4. R2 statistic measures the proportion of variability in response that can be explained using

predictors. True or False.

5. Training set RSS and training set R2 cannot be used to select from a set of models with

different numbers of predictors. True or False.

6. Name one algorithm used to perform simple linear regression. What is the quantity optimized?

Y = β0 + β1x Ordinary List Squares

RSS = ∑(y-ӯ)^2 RSS / dβ0 = 0 RSS / dβ1 = 0

7. What are the two main goals of data-driven modeling?

Prediction and Information.

8. Name one approach for estimating test set error. Explain the approach in 2-3 sentences.

Live one out cross , validation set approach(do multipule times) randomly devide training set.how it
works.
9. Answer the questions based on what is provided in the given box. Consider a statistical

significance level of 0.05 for this problem.


(KESİN SORACAK MODELLER FARKLI OLACAK)

Alfa = 0.05 (cut-off)

a) Describe the null hypothesis for the F-test?


None of the predictors (Age , weight, height, pulse , diastol) has an effect on systol.
b) Is there a relationship between the predictors and the response? Justify your answer.
Go check the F test. Check the p-value. Its smaller than our cut-off. Since its smaller we reject
the null. There is a relationship between the predictors and the response.
c) Describe the null hypotheses, which the p-value for Weight corresponds to.
Weight has no effect on Systol in the presence of other predictors in this model. Because it’s
a multiple linear regression.
d) Which predictors appear to have a statistically significant relationship to the response?
Justify your answer.
Weight has statistically significant effect with the Systol. Because its p-value is smaller than
our cut-off.
10. Hasan works for a pharmaceutical company and conducted a study to assess the efficacy and
potency of a group of drugs before allowing the agents to proceed to clinical trials. Taking the Hill
coefficient as 1, he fitted a Hill model. Figure given below shows the normalized doseresponse curves
he fitted for four different drugs from the same class of medications (Drugs A, B, C, & D).

Efficacy: maximum effect that a drug can produce regardless of dose. (ALL DUGS HAVE EQUAL
EFFİCACY)

Potency: amount of a drug that is needed to produce a given effect (e.g, for instance EC50). (LOW
CONSENTRATİON İS GOOD)

Which of the following is true regarding these medications?

A. Drug A has greater efficacy than Drug D.

B. Drug D has greater efficacy than Drug A.

C. Drug A has greater potency than Drug D.

D. Drug D has greater potency than Drug A.

E. All four drugs (A,B,C,D) have equal potencies.


11. Barış works for a pharmaceutical company and he is supposed to assess the efficacy and

potency of 2 drugs before allowing the agents (Agent A and Agent B) to proceed to clinical

trials. He performed a preliminary experiment to get an overall idea and obtained some data.

He decided to model the dose-response relationship using a Hill Model in MATLAB (see

below). Alfa = 1
% The data B is parameter
>> dose=[1;10;60;100;200;540;600;850;1000]; % (mg/kg) X is concentration
>> response_A=[8;43;70;74;77;78;79;79;85]; % (% response) E = Emax C / EC50 +C
>> response_B=[9;49;84;89;93;95;96;96;98]; % (% response) b = [Emax
% Define the Hill function x is the concentration EC50]
>> fitHillModel=@(b,x) b(1).*x./(b(2)+x); Answer for a
% Define initial values

>> init_A=[max(response_A);median(dose)];

>> init_B=[max(response_B);median(dose)];

% Fit the model using defined function

% and the MATLAB lsqcurvefit() function.

>> mdl_1=lsqcurvefit(fitHillModel,init_A,dose,response_A);

>> mdl_1

mdl_1 = Emax (bcs its first entrie)

81.0451

9.0467 EC50

>> mdl_2=lsqcurvefit(fitHillModel,init_B,dose,response_B);

>> mdl_2

mdl_2 =

97.6532 Emax

9.8931 EC50

a) Write the form of the Hill Equation he used. State the variables and the parameters the equation
based on.
b) What are the values for the Hill slope (α), Emax and EC50 for each drug?

Hill slope (α) = 1

For A Emax = 81.0451 For A EC50 = 9.0467

B Emax = 97.6532 B EC50 = 9.8931

c) Compare the drugs based on their efficacy and potency. Which one would you prefer? Explain

for potency we check EC50. high potency means low EC50. High efficacy high Emax.

We want high potency and high efficacy

their potency is the same. Very close.

Drug 2 is better due to its higher efficacy.

You might also like