ECON121 L1 , L2:

Tags

Lecture 1
X = Random Variable
θ ^ = Estimator
θ = Parameter

3 Important Properties of an Estimator:⁉️

Unbiased = Expected Value of Estimator = True Value of Parameter; i.e Bias =

0
Consistent = As we increase sample size, θ^ → θ (θ^ converges in probability to
θ)

Efficient = θ^ Has the least possible variance.

Cannot have all 3 Properties: ⁉️

Either 1 and 2 ( Unbiased and Consistent)

Or 2 and 3 (Consistent and Efficient)

ECON121 L1 , L2: 1
 If we have Unbiased(1) and Consistent(2) but very INEFFICIENT(3) (variance is
too high) then θ^ is not very useful as it is very noisy

Standard Deviation vs Standard Error:⁉️

Standard Deviation is used for a known variable X.

Standard Error is used for a computed Estimator θ^.

Lecture 2: Means , T-Tests & Regressions:

μ - Population Mean
X-bar = Sample Mean

1) Sample Mean

X-bar is the mean of the sample, hence X-bar can change as we draw different
samples.
Therefore we can say X-bar = μ^ as it is an Estimator of μ.

Properties of Sample Average:

BLUE : We can say that X-bar is the Best Linear Unbiased Estimator of μ.
This means that for all possible estimators μ^, X-bar is the least-biased. E[
X-bar] = μ.

ECON121 L1 , L2: 2
 Proof:

Law of Large Numbers: As N, the size of the sample increases the value of
X-bar approaches the population mean- μ. (X-bar → μ). i.e Consistency
Principle

Central Limit Theorem: As N, the size of the sample increases, the

distribution of our sample mean X-bar approaches a Normal Distribution.

This means that E[ X-bar] = μ & V[ X-bar] = σ2/N

Proof:

2) Sample Variance

We know that Sample Mean, X-bar is the Best Linear Unbiased Estimator of μ.

But Sample Variance is NOT the the BLUE of the Population Variance, σ2
Proof:

ECON121 L1 , L2: 3
 Expected Value of the estimator Sample Variance does NOT equal the
Population Variance. E[σ2^] =/= σ2 .

Hence Sample Variance (σ2^) is not Unbiased. ⁉️

2.1) Degrees of Freedom Adjustment:

We already proved:
Sample Variance (σ2^) is Biased and E[σ2^] =/= σ2

The real Expected value of σ2^ is

So how do we make it Unbiased?

We use this instead:

ECON121 L1 , L2: 4
 Full proof for above:
https://economictheoryblog.com/2012/06/28/latexlatexs2/

Hence we have 2 Estimates of E[σ2^]:

The only difference being in the [1/(N-1)] vs [1/N].

We notice that Subtracting 1, results in an Unbiased Estimator but with

Higher Variance

Keeping it as [1/N] results in a Biased Estimator but with Lower Variance

The above situation represents 2 important concepts⁉️:

1) Degrees of Freedom Adjustment: Subtracting 1 from the sample-size

N is 1-degree of freedom adjustment.
2) Bias-Variance Tradeoff: We cannot reduce the Bias without reducing
Variance and vice-versa. So we have to chose appropriately

*** It is worth noting, as N increases the bias term reduces and we

can overcome the tradeoff

3) T- Testing:
We now know ways to find Sample Means (X-bar) and Sample Variances.

But these Sample Means change from sample to sample. So, How to
compare Means of different Samples ⁉️ - Using T-Testing

ECON121 L1 , L2: 5
 A T-Test is a statistical test that is used to compare the means of two
groups. It is often used in hypothesis testing to determine whether a
process or treatment actually has an effect on the population of interest, or
whether two groups are different from one another.

Involves computing a t-statistic :

θ^ refers to the estimator being tested for - usually the Sample Mean.

θo refers to the Null Statistic/ Null Hypothesis - usually Zero [How

different from 0 is θ^?]

SE[θ^] refers to Standard Error of θ^ - equals 1 if sample size high,

Central Limit Theorem

1-Sample t-Test:
Used when we have only one sample to compare. Usually compared
against 0

Null Hypothesis: X-bar = 0

ECON121 L1 , L2: 6
 5 Confidence Level == 0.05 p-value == +1.96 to -1.96 Z-Score == 95%
Confidence Interval

We reject the Null Hypothesis if t-stat > 1.96 or < - 1.96‼️⁉️

2-Sample t-Test:
Used when comparing sample means across different samples and
populations.

Eg: Want to compare mean income between Blacks and Whites. How
to see if they are significantly different from each other?

95% Confidence Level == 0.05 p-value == +1.96 to -1.96 Z-Score

We reject the Null Hypothesis if t-stat > 1.96 or < - 1.96‼️⁉️

4) Confidence Intervals:
For a given Confidence Level, say 95%. The confidence interval will contain
the parameter θ in 95 out every 100 samples.

If the hypothesized value falls outside the interval, we reject the null
hypothesis.

5) OLS Regression:
We want to observe how 2 variables, X and Y behave in reference to each-
other:

ECON121 L1 , L2: 7
 When we have samples and estimated functions, the equation above changes
to:

In Equation 1, we call the difference between Y and the Function as - Error

In Equation 2, we call the difference between Yi and the Function as - Residual

Solutions to OLS Regression:

The ‘solution’ refers to those values of b0^ and b1^ that minimize the loss
function
We call these β0 and β1 respectively

Empirically their values can be calculated as:

ECON121 L1 , L2: 8