lOMoARcPSD|44031930

Cheat Sheet Compilation

Financial Modelling (Western Sydney University)

Scan to open on Studocu

Studocu is not sponsored or endorsed by any college or university

Downloaded by mit daz ([email protected])
 lOMoARcPSD|44031930

Econ 103 Cheat Sheet

Old final questions - Atkins Final Spring 2015 - Watchter Final Spring 2014
1) If x and y are ​complements​, an increase in the price of x will lead to a decrease in the quantity demanded of y. If x and y
are ​substitutes​, an increase in the price of fx will lead to an increase in the quantity demanded of y. Look at the sign of
coefficients of relevant variables to make an assessment.
2) SSE​k≥SSE​
​ k+1​ ​always

3) The large constant estimate in both regressions suggests that ​there is no implication regarding the impact of the
other variables.
4) Which of the following statements is true regarding var(​B^​ ​2​) when estimated by IV using ​z​ as an instrument for ​x​?
Instrumental variable estimation leads to larger variance of estimates compared to OLS​.
5) What is the null hypothesis when performing an ​F​-test to test the strength of multiple instruments ​j​ = 1,...,J? ​The
instruments are weak with all coefficients θ​j​ on the instruments different from 0.
6) In certain country, most people are poor but some people are very rich. Let X to be the wealth of an individual randomly
chose from the population of that country. Then, ​the mean of x is greater than the median of X. In addition, the
skewness of X is positive.
7) When you are implementing an instrumental variable regression, you are worried about
a) a potential direct effect of the IV on outcome

m
er as
b) a weak relationship b/n IV and endogenous variable
c) a remaining correlation of IV and error term

co
8) The interpretation of slope coefficient β​k​ in the model is ​a change in x​ki​ by one unit is associated with a β​k​ change in

eH w
Y, holding all the K-1 regressors constant.

o.
9) A high R​2​ or adjusted-R​2​ does not always mean that an added variable is statistically significant.

rs e
10) Omitting a variable which is relevant can result in a negative value for the coefficient of the included variable, even
ou urc
though the coefficient will have a significant positive effect on Y if the omitted variable were included.
11) As sample size N increases, the length of 100(1 - α) confidence interval ​decreases.
12) One will reject H​0​: β​k ​= 0 against H​1​: β​k​ > 0 at significant level α ​the 100(1 - α) cannot help in testing this
hypothesis​.
o

13) SST is measured differently for ​y = β​0​ + β​1x​ ​ 1​ + e and ln(y) = β​0​ + β​1x​ ​ 1​ + e.
aC s

14) To test whether or not the population regression is linear rather than a polynomial of order r, ​use the test of (r-1)
vi y re

restrictions using F-statistic.

15) The binary variable interaction regression ​allows the effect of changing one of the binary independent variables to
depend on the value of the other binary variable.
16) In the regression model y = β​0​ + β​1x​ ​ 1​ + β​2D​
​ i​ + β​3(X​
​ iD​
​ i)​ + u​i,​ where x​i is a continuous variable and D​i​ is a binary variable,
ed d

​
to test that the effect of x​i​ on y is identical for both values of D​i​ you must use ​a t-test for H​0:​ β​3​ = 0.
ar stu

17) When an exogenous variable is used, IV estimators are ​consistent and approximately normally distributed in large
samples.
18) What is the null hypothesis when performing an F-test to test the strength of multiple instruments j = 1,..., J?​ T ​ he
instruments are weak, with no coefficient θ​j​ on the instruments in the first stage different from 0.
is

19) ​TestScore​i​ = ​607.3 + 3.85​Income​i​ - ​0.0423​Income​i2​ ​. The equation ​suggests a positive relationship b/n test scores
and income until a value of the income of approximately 45.508.
Th

20) ln(WAGE) = 1.439 + 0.0834EDU + 0.0512EXPER + 0.1932WHITE. ​White employees earn 19.32% more than
non-whites.
The log-linear regression equation can be interpreted such that a one unit increase in an x​k​ variable (independent
sh

variable) leads to approximately a (100)β​k% ​ change in the y (dependent) variable.

21) The Chow test i​s an F-test for the equivalence of two regressions.

Buchinsky 2011
● model that uses income to predict monthly expenditures on transportation --​ explanatory variable: income
● NOT assumption of SLR Model - ​parameter estimate of B2 is unbiased.
● OLS model, N increases, variance b2 ​decreases
● MR model: which does NOT lead to larger variances of least square estimator b​2​ and var(b​2​) - ​larger correlation
between x​2​ and y
● degrees of freedom in denominator of F-distribution - ​the number of observations minus number of coefficients
estimated (N-K)
https://www.coursehero.com/file/12911867/CheatSheetCompilation/
1
Downloaded by mit daz ([email protected])
 lOMoARcPSD|44031930

● how does omitting relevant variable from regression model affect estimated coefficient of other variables - ​biased, can
be positive or negative
● when collinear variables are included in an econometric model coefficient estimates are ​unbiased but have larger
standard errors
● Running auxiliary regressions where each explanatory variable is estimated as a function of the remaining explanatory
variables can help detect ​collinearity
● adjusted R^2 is better measure than R^2 because ​it adjust R^2 for the number of variables in the regression
● log log regression - corr​x2, x3​ = 0; ​such model cannot be estimated by OLS
● model with B3>0, test B2/B3 = 1-- then ​you can use t-test (or F-test) considering the null H​0:​ B2-B3=0.
● Gauss Markov does NOT depend on assumption: ​values of e are normally distributed
● any given linear model, let b1 be the OLS estimator of B1. ​b1 is a random variable, whereas B1 not.
● country w mostly poor and some very rich. ​mean of X is greater than median. Skewness is POSITIVE
● if male and female dummy variables in a regression: ​regression cannot be estimated due to perfect collinearity
● police regression: ​ in order to measure this impact we need to use an instrumental variable
● regression of one more year of education on the wages for blacks: ​effect given by B2 + B4.
● restricted model has smaller R^2 compared to the original model
● randomized, controlled experiments are needed to accurately measure treatment effects without ​omitted-variable bias

m
Atkins - Practice Final Spring 2015

er as
● The interpretation of the slope coeff B​k​ is - a change in x​ki​ by one unit is associated with a Bk change in Y, holding all

co
other K-1 regressors constant

eH w
● R​2 ​is a valid measure of goodness of fit of the regression if - the regression has a constant term
False - a high R​2​ or Rbar​2​ always means that an added variable is statistically significant

o.
●
●
rs e
If you had a two regressor regression model, then omitting one variable which is relevant → can result in a neg value for
the coefficient of the included variable, even though the coefficient will have a significant positive effect on Y if the
ou urc
omitted variable were included
● When collinear variables are included in an econometric model, coefficient estimates are → unbiased but they have
larger standard errors
o

● If one rejects the null hypoth H​0​ = B​k​ = 0 against H​1​ = B​x≄​ 0 at the significant level alpha then → it cannot be determined
aC s

whether he/she will reject it for H​0​ = B​k​ = 0 against H​1​ = B​k​ > 0
vi y re

● To decide whether (linear reg) or (log regression) fits the data better, you cannot consult the regression R​2 ​because →
the SST are not measured in the same units between the two models
● To test whether or not the population regression function is linear rather than a polynomial of order r → use the test of
(r-1) restrictions using the F-statistic
ed d

● The binary variable interaction regression → allows the effect of changing one of the binary independent variables to
ar stu

depend on the value of the other binary variable

● When an exogenous instrument is used, IV estimators are → consistent and approx normally distributed in large
samples
● True regarding var(B​2)​ when estimated by IV using z as an instrument for x → instrumental variable estimation leads to a
is

larger variance of estimates compared to OLS

● What is the null hypoth when performing an F-test to test the strength of multiple instruments j=1...J → the instruments
Th

are weak with no coeff theta on the instruments in the first stage different from 0
● when you are implementing an instrumental variable regression, you are worried about → all of the following = a
potential direct effect of the instrumental variable on the outcome; a weak relationship between the instrumental
variable and the endogenous variable; a remaining correlation of the instrumental variable and the error term
sh

Watchter - Practice Final March 2014 (almost identical to above exam)

● Choosing the right functional form for the regression is important because → it allows one to capture the specific
features of the observed data

Buchinsky - Dec 2014

● Following leads to larger forecast errors → ​the value of (x​0​ - xbar)​2​ is larger
● How does omitting a relevant variable from a regression model affect the estimated coefficient of other variables in the
model →​ they are biased and the bias can be negative or positive

https://www.coursehero.com/file/12911867/CheatSheetCompilation/
2
Downloaded by mit daz ([email protected])
 lOMoARcPSD|44031930

● Running aux regressions where each explanatory variable is estimated as a function of the remaining explanatory
variables can help detect → ​collinearity
● The adjusted R​2​ Rbar​2 ​is a better measure than R​2​ because → it adjusts R​2​ for the number of variables in the regression
● In a simple regression model, the Gauss-Markov Th does not depend on → ​the values of e are normally distributed
● Which cannot cause the least square estimator to be biased → ​heteroskedastic random error variance^2 is not
constant across different observations
● In a multiple regression, if we change the units of measurement of y by multiplying it by some constant, estimator of B​0​,
B​1,​ B​2​ and standard errors of B​0,​ B​1,​ B​2​ would change → ​R2​ ​ would not change
● With a log model →​ if we increase 1% of x​i,​ y​i​ increases B​1​ percent
● With a full model and a restricted model → ​R2​ ​ of the full model is always greater than that of the restricted model
● Under heteroskedasticity, the OLS estimators → ​are still linear but no longer attain the minimum value (still
unbiased)

Buchinsky - Dec 2013

● Suppose you are interested in the hypothesis H​0​ = 2B​2​ -B​3​ = 0 against H​1​ = 2B​2​ - B​3​ ≠ 0. To test this hypoth, one needs
to know →​ the estimates for B​2​ and B​3​ and their variance-covariance matrix
● Suppose that Var(e​i)​ = h(x​2​). Then →​ the variances, of all estimates b1...bk need to be corrected for this fact
● If we were to omit x​K​ from the regression from the regression then → ​the coefficients for the other variables will not

m
be biased if x​K​ is not correlated with x​2...x​

er as
​ K-1

● One will reject the null hypothesis H​0​ = B​l​ = 0 against H1 : Bk > 0 at the significance level a →​ the 100(1-a) confidence

co
interval cannot help in testing this hypothesis

eH w
● Suppose that Var(ei) = o-^2. Then the Gauss-Markov th → ​does not apply to the least squares estimator B1... BK
If the x’s in the regression x1….xk are uncorrelated, then → ​we cannot determine the sign for Cov(bk, b1) for all k ≠

o.
●
l, k,l = 1...K
rs e
In a given data set the larger the R​2​ →​ the lower is SSE
ou urc
●
● When collinear variables are included in an econometric model, coefficient estimates are → ​unbiased but they have
larger standard errors
● Choosing the right functional form for the regression is important b/c → ​it allows one to capture the specific feature
o

of the observed data

aC s
vi y re

Hypothesis Testing
● In one rejects H​o:​ B​k​ = 0 for all k = 2,...,K against H​1:​ B​k​ ≄ 0, for some at least k, at the significance level α​1,​ then he ​will
reject it for all α​2​ > α​1
● If we were to reject the H​0​: B​k​ = 0 against H​1​: B​k ​ ​≄ 0 for some k, k = 2,...,k then ​we are likely to reject H​0​: B​2​ = … = B​k​ =
ed d

0 against H​1​: Not H​0

ar stu

● Consider the case in which K = 3, let x​2i​ = 1 if the person is a male and zeros otherwise, and x​3i​ = 1 if the person lives in
the LA area and zero otherwise, while y​i​ is the individual’s earnings. The average earnings for a female living in LA is ​B​1​ +
B​3
● If one rejects H​0​: B​k​ = 0 against H​1​: B​k​ ≄ 0 at significance level α, then ​it cannot be determined whether he will reject
is

it for H​0​: B​k ​= 0 against H​1​: B​k ​≄​ 0

Th

Assumptions
Assumptions of Simple Linear Regression Model
SR1: y = β​1​ + β​2x​ ​ 2​ + e
sh

SR2: E(e) = 0 or E(y) = β​1​ + β​2x ​

SR3: var(e) = σ​2​ = var(y)
SR4: cov(e​i​, e​j​) = cov(y​i​, y​j​) = 0 → the random errors ​e​ are statistically independent
SR5: The variable ​x​ is not random and takes on at least two different values
SR6: (optional) e is normally distributed

Assumptions of Multiple Regression Model

MR1: y​i​ = β​1​ + β​2​x​i2​ +...+ β​k​x​ik​ + e​i
MR2: E(e​i​) = 0
MR3: var(e​i)​ = σ​2​ → existence of heteroskedasticity will violate this assumption
MR4: cov(e​i,​ e​j)​ = 0 for any i ≠ j
https://www.coursehero.com/file/12911867/CheatSheetCompilation/
3
Downloaded by mit daz ([email protected])
 lOMoARcPSD|44031930

MR5: each x​ki​ is nonrandom, and is not exact linear functions of other explanatory variables
MR6: (optional) e​i​ is normally distributed

Properties of the IV Estimator

They are consistent, if z is exogenous, with E(ze) = 0
In large samples the instrumental variable estimators have approximate normal
distributions
as well ​→

Math-Based Questions

● Elasticity: dy/dx x x/y

● R^2 = r^2xy = (sxy/sx x sy)^2
● Solving for estimate b1 =
○ y = b1 + b2x2 +b3x3
● Variance (a1b1 + a2b2) = a1^2var(b1) + a2^2var(b2) + 2a1a2 cov(b1,b2)

m
er as
○ SE = sq rt of variance
● To find the point at which a differentiable function changes slope, take the first derivative and set it equal to 0

co
eH w
○ dTestscore/dIncome = a + bincome = 0

o.
● T-stat = λ - c / se (predicted value λ)

rs e
○ se (predicted value λ) is always positive so the sign of predicted value λ determines the sign of
ou urc
the t-stat
○ se (predicted λ) = sq variance (predicted λ)
○ from there, make se (predicted λ) the denominator and solve for t-stat formula
o

● Definition of F-test statistic = SSE1 - SSE2 / SSE2 x N-K/1

● R1^2 = 1 - SSE1/SST and R2^2 = 1 - SSE2/SST
aC s

○ R2^2 - R1^2 = SSE1 - SSE2 / SST

vi y re

○ 1 - R2^2 = SSE2/SST
○ R2^2 - R1^2/1 - R2^2 = SSE1 - SSE2 / SSE2
● SST = SSE + SSR
● SSE = SST - SSR
ed d

● F = SSR2 -SSR1/SST - SSR2 x N-K/1

ar stu

● For output, elasticity = B2 x x/y

Labs
10/2/15​: ​Regression Analysis in Employment Litigation – Elaine Reardon
is

● Resolution Economics
Th

● Discrimination, work breaks, etc.

● Defining “similarity”
● Employment discrimination is itself not observable
● ln(hourly pay)=cons_+a​1(hispanic)+b​
​ 1(explanatory_variables)+error
​
sh

○ If a​1​ is negative then we would not observe discrimination based on the data
● Pay gap is statistically significantly lower for Hispanics
● Too many variables that aren’t justifiable?
● A​nalyzes complex data for the purpose of assisting counsel in evaluating class certification and merits issues in
employment matters. She has analyzed claims alleging age, race, and gender discrimination in hiring, termination,
and pay equity. She also has considerable experience in wage and hour consulting and litigation, utilizing various
data sources such as surveys, observation studies, and administrative data to assess class certification issues
regarding uncompensated time, missed meals, and exempt/non-exempt status. She has significant experience in
designing, implementing and analyzing scientific surveys, including drawing statistical samples and making
inferences from the results. In connection with her litigation work, Dr. Reardon has served in an expert capacity a
number of times.
https://www.coursehero.com/file/12911867/CheatSheetCompilation/
4
Downloaded by mit daz ([email protected])
 lOMoARcPSD|44031930

10/9/15: Econometrics Careers Explored – JJ Espinoza

● Credit card fraud – standard deviation of normal transactions
● Determine how Walt Disney Company should structure advertising in the course of the introduction of new movies
10/16/15: The Effects of Regulated Marijuana Market Growth on Adult and Adolescent Substance Use – Rosana
Smart
● Ogden removed federal risk
10/23/15: Consulting & Economics – Marshall Crockett
● PwC
● Forensics
10/30/15: Commercial Real Estate – Kidder Matthews
● Residential multi-units, property management, etc.
● forecasting and trending for acquisitions, dispositions and debt restructure

11/6/15: RCLCO Real Estate Advisors – Taylor Mammen

● Strategic guidance for complex, multi-use, large-scale land use challenges, including master planned communities
and “new cities,” urban mixed-use developments, and redevelopment/ revitalization strategies
11/13/15: Marketing Science at Lieberman Research Worldwide – Hillary Decamp
● Los Angeles Times’ readership → 10 pt drop in readership for each successive generation

m
● Customer churn

er as
● Targeting potential customers

co
○ Segmentation

eH w
○ Factor analysis to search for redundancies
● Garbage in garbage out (GIGO)

o.
● Turf analysis
rs e
ou urc
Corn Production
● The coefficient B​2​: a unit increase in capital leads to a 265% increase in the production of corn
● According to the above result, we reject H​0:​ B​3​ - B​4​ = 0 against H​1:​ B​3​ - B​4​ ≄ 0: at 1% and 5%
o

● According to the above result, we reject H​0:​ B​4​ = B​5​ = 0 against H​1:​ B​4 ≄​ 0 or B​5​ ≄ 0: there is not enough information
aC s

● Considering the above output results, the consequence of heteroskedasticity in e is: all of the above
vi y re

Wage regression for Rich and Poor Output

● If we were to take im_rich out of the regressor, but include native, the estimate for im_poor would now be: between
-0.30 and -0.15.
If we were to take im_rich out of the regressors but include native, the R​2​ would now be: equal to the one of the
ed d

●
original model.
ar stu

● Consider the original model. Suppose that we think that the wage has a higher variability among immigrants than
among natives. Then: the OLS estimator would still be unbiased . The estimated intervals would not be valid.
● We think that education has a positive effect in wages, but immigrants from poor countries are on average less
is

educated than natives. We conclude that: the OLS estimator of im_poor has a negative bias. We expect B​2​ to be
greater than -0.1009196
Th

● if we had measured the wages in cents instead of dollars: the constant would be larger, the coefficient of im_rich
would be the same.

Mroz Data
sh

● Section 2 results: ​one needs to adjust for potential heteroskedasticity

● correlation matrix in Section 1: ​the regression results are valid
● Section 3 results indicate that once one accounts for potential heteroskedasticity: ​coefficient on the wife’s
experience becomes insignificant at the 1% level
● education and experience of the husband are negatively correlated, while that for the wife is positively
correlated
● living in large city: ​total family income is higher
● husbands tend to have more labor market experience than their wives
● Section 3 results: ​the coefficients on the husband’s experience and the wife’s experience are jointly
significant

https://www.coursehero.com/file/12911867/CheatSheetCompilation/
5
Downloaded by mit daz ([email protected])
 lOMoARcPSD|44031930

● Section 4 results: ​total effect of husband’s and wife’s experience is insignificant from zeros at the a =.01
significance level
● 95% CI: ​(25,618, 41,539)
● having 3 children under 6 years old: ​no significant statistical effect on family income​.

Demand for Meat

● The F-stat for H​0​ - B​2​ = B​3​ = B​4 ​= 0 from the first regression is: 1.70
● The results indicate that: it is essential to have the variable ​income ​in the regression
● The results of the first regression indicate that the hypothesis H​0:​ B​2​ = B​3​ = B​4​ = 0
● The point estimates from the second regression indicate that: meat and vegetables and fruit are substitutes, while
meat and bread and cereal are complements
● Based on the results of the second regression, the t-stat for the test of H​0​: B​3​ + B​4​ = 0 against H​1​: B​3​ + B​4 ​≄ 0 will
be -0.6785
● Consider testing the restriction in model that B​5​ = 0: all of the above
● The large constant estimate in both regressions suggests that: there is no implication regarding the impact of the
other variables
● The interval estimate for the coefficient on income, i.e., B​5​, for α = 0.05 is approximately: [+1.566, +2.912}
● The fact that R​2​ in model 2 is substantially higher than in model 1 means: that income explains an important part of

m
the variance of meat purchases

er as
● The fact that coefficients on prices in model 2 barely change as income is included as regressor implies: that there

co
is a low correlation of individual income and these prices

eH w
● The elasticity of meat purchases with respect to a change in price in meat: cannot be calculated without further
information

o.
● The 95% confidence interval for the effect of price of meat on meat purchases: is approximately [-1.1629, -0.1885]
rs e
ou urc
Boca Raton Output
● The results indicate that a traditional house would cost approx → ​$18k
● The correlation between the coefficients on the number of bedroom and the number of baths = ​-0.01835
o

● The results indicate that if one test the hypothesis H0:B2 =...Bk = 0 against H1: at least one Bk ≠0, k = 2...K one
→ ​will reject the null hypoth a = 0.01
aC s

Holding all variables constant, the result indicate that, on avg, having a pool → ​does not have an effect on the
vi y re

●
house price
● The 95% CI for the coefficient on SQFT is approx → ​[74.46,89.20]
● The houses have waterfront → ​7.2% of the houses have waterfront
The results indicate that the effect having a traditional house with a pool is → ​negative
ed d

●
ar stu

Wage Regression Output

● The variable NorthEast is dropped because of → ​collinearity
● Individuals in MidWest earn on average 4.232 less than →​ individuals in NorthEast
is

● The results indicate that if one test the null hypoth H0 = B3 = B5 = B6 = 0 against H1: at least one Bk ≠ 0, k =
3,5,6 → ​one will reject the null hypothesis for a = 0.1
Th

● Consider the rejected model with the form WAGE = B1 + B2EDU → ​the restricted model will have a smaller
R^2 compared to the original model
sh

https://www.coursehero.com/file/12911867/CheatSheetCompilation/
6
Downloaded by mit daz ([email protected])
Powered by TCPDF (www.tcpdf.org)