Eco311 Optional Reading: Standard Error and Confidence Interval
(Jing Li, Miami University)
1. This note uses Monte Carlo simulation to help students understand concepts of stan-
dard error and confidence interval.
2. Statistics is about using samples to understand population. An important fact is that
there are many samples for a given population. For instance, the population can be
all students at Miami university. Then students taking a class in Laws Hall room
304 at 3pm on Monday can be a sample. Another sample can be students taking a
class in room FSB 0019, or students eating at Chipotle. Most likely, we get different
results from different samples. A key issue in statistics is accounting for the variation
or uncertainty in those different estimates.
3. Recall the math we did in class: if you obtain many random or iid samples, and
compute many sample means, then the variance of those many sample means is
var(¯y) =
σ
2
n
(1)
where σ
2
= var(y) is the variance of data. The square root of var(¯y) is standard error
se, while the square root of var(y) is standard deviation sd:
se =
r
σ
2
n
=
σ
√
n
(2)
sd =
√
σ
2
= σ (3)
The former measures variation in ¯y, while the latter measures the variation in y. Do
not confuse them.
4. From (2) it is obvious that as the sample size n rises, the standard error
σ
√
n
falls.
This fact implies that as samples get larger, the variation in ¯y across different samples
gets smaller, or equivalently, ¯y becomes preciser. In light of this, the standard error
can serve as a measurement of uncertainty or unpreciseness of sample estimates. We
prefer precise estimates or small standard error. We can achieve that goal by using
large samples.
5. Next we run Monte Carlo simulation, and create a sample of 1000 observations of
1