Hypothesis Testing with Large Samples: Switching from t‑Distribution to Z‑Distribution

Name: 01 - Hypothesis Testing For Means & Large Samples, Part 1
Uploaded: 2026-02-17T04:41:59.278499+00:00
Channel: Math and Science
Description: Summary and key takeaways on Hypothesis Testing with Large Samples: Switching from t‑Distribution to Z‑Distribution, covering Introduction In this lesson we

Math and Science

Feb 17, 2026

•

3 min read

YouTube video ID: 80YzzIm8NK8

Source: YouTube video by Math and Science — Watch original video

PDF

Introduction

In this lesson we move from hypothesis testing with small samples (n ≤ 30) to testing with large samples (n > 30). The core steps of hypothesis testing stay the same, but the underlying probability model changes from the t‑distribution to the normal (Z) distribution.

Why Sample Size Matters

Small samples (n ≤ 30) – use the t‑distribution because its shape depends on the degrees of freedom (df = n‑1). With few observations the curve is wider and heavier‑tailed.
Large samples (n > 30) – the t‑distribution becomes virtually indistinguishable from the normal distribution. Therefore we adopt the standard normal (Z) distribution, which has a fixed bell shape regardless of sample size.

The Test Statistic for Large Samples

The Z‑statistic is calculated exactly as the t‑statistic, only the symbol changes:

Z = (X̄ – μ) / (s / √n)

where: - X̄ = sample mean - μ = hypothesised population mean - s = sample standard deviation - n = sample size

Steps of Hypothesis Testing (unchanged)

State hypotheses (null H₀ and alternative H₁).
Choose significance level α (e.g., 0.05).
Determine the rejection region using critical Z values.
Compute the Z statistic from the data.
Compare the statistic to the critical value(s) and reject or fail to reject H₀.

Critical Z Values for Common Confidence Levels

Confidence	α (two‑tail)	One‑tail Z	Two‑tail Z (±)
90 %	0.10	1.28	±1.645
95 %	0.05	1.645	±1.96
98 %	0.02	2.05	±2.33
99 %	0.01	2.33	±2.575

One‑tail test: use the positive Z for a right‑tail test, the negative Z for a left‑tail test.
Two‑tail test: use both the positive and negative Z values; each tail contains α/2 of the total area.

Advantages of Using the Normal Distribution

Fixed shape: No need to look up different tables for each df; the same critical values apply to any n > 30.
Simplified calculations: Once you have the table of Z‑values (or a calculator), you can reuse them across problems.
Consistent tabulation: Normal tables give the area to the left of Z, whereas t‑tables give the area to the right of t. Keep this orientation in mind when you switch between them.

Practical Tips

Memorise the most common Z critical values (1.28, 1.645, 1.96, 2.05, 2.33, 2.575).
When the problem states a confidence level, instantly translate it to α and pick the corresponding Z.
Always verify whether the test is one‑tailed or two‑tailed before selecting the sign of the critical value.
Use a scientific calculator or statistical software to compute Z quickly, but understand the manual lookup process for exam settings.

Summary

For sample sizes larger than 30, hypothesis testing proceeds exactly as before, except that the t‑distribution is replaced by the standard normal distribution. This change eliminates the need to adjust critical values for different degrees of freedom, making the procedure faster and less error‑prone.

When you have 30 or more observations, switch to the Z‑distribution; the test steps stay the same, but critical values are fixed, simplifying hypothesis testing dramatically.

Frequently Asked Questions

Who is Math and Science on YouTube?

Math and Science is a YouTube channel that publishes videos on a range of topics. Browse more summaries from this channel below.

Does this page include the full transcript of the video?

Yes, the full transcript for this video is available on this page. Click 'Show transcript' in the sidebar to read it.

Why Sample Size Matters

- **Small samples (n ≤ 30)** – use the **t‑distribution** because its shape depends on the degrees of freedom (df = n‑1). With few observations the curve is wider and heavier‑tailed. - **Large samples (n > 30)** – the t‑distribution becomes virtually indistinguishable from the normal distribution. Therefore we adopt the **standard normal (Z) distribution**, which has a fixed bell shape regardless of sample size.

Helpful resources related to this video

If you want to practice or explore the concepts discussed in the video, these commonly used tools may help.

Statistics For Dummies Paperback Recommended

A beginner‑friendly guide that explains core concepts like hypothesis testing and Z‑scores, helping readers apply the lesson immediately.

Amazon →

The Art Of Statistics Book

Provides deeper insight into statistical thinking and real‑world applications of confidence intervals and hypothesis tests.

Amazon →

Scientific Calculator Handheld

Allows quick computation of Z‑statistics and other formulas without needing a computer, essential for exam settings.

Amazon →

Z Score Calculator Handheld

A dedicated device that instantly converts raw scores to Z‑values, reinforcing the normal‑distribution method taught in the lesson.

Amazon →

Statistical Tables Book

Contains pre‑printed Z‑critical values for common confidence levels, eliminating the need to search online during tests.

Amazon →

Links may be affiliate links. We only include resources that are genuinely relevant to the topic.

Summarize another video

Full Transcript YouTube

hello welcome to this lesson in
mastering statistics here we're going to
work on hypothesis testing but in a
slightly different case and that's kind
of how the rest of this uh set of
lessons is going to go we're going we
have already done hypothesis testing
we're doing testing on means when the
sample size that we have collected for
our testing is less than 30 30 or less
okay now we're getting into a situation
now what we did before was what we
called small sampling when we have less
than 30 samples here we're doing what we
call large samples which as you might
guess is greater than 30 uh samples and
this very very closely follows what
we've done when we did confidence
intervals we also had small uh samples
and large samples it's the same sort of
thing so we use 30 as our kind of cut
off below that we use the T distribution
that we've been using all this time and
then here now that we're getting into 30
samples or greater or greater than 30
samples we're going to be doing the
normal distribution now if you remember
back a long time
ago uh I taught you uh you may not
remember but I taught you that the T
distribution it looks bell-shaped but
the exact shape of it depends on the
degrees of freedom and that depends on
the number of samples remember degrees
of freedom is the number of samples
minus one n minus one so if you have a a
very small amount of samples the curve
is going to also look bell-shaped but
maybe not quite the same as it does if
you have 30 or 40 samples it turns out
that as you have more and more and more
samples for a t distribution it looks
more and more and more normal so T
distribution looks like a normal
distribution when you have a large
number of samples so that's why when we
had what we call small samples we had to
actually use the T distribution because
we didn't have enough samples so that
the T distribution looks normal enough
to look to use the normal distribution I
know that's a lot of words but bottom
line is when we have a small number of
samples and the difference between a t
distribution and a normal distribution
is pretty pretty great difference so we
have to use the full-blown T
distribution when we get over 30 samples
it's okay to use a normal distribtion
tion because a t distribution looks like
a normal distribution when you have a
large enough number of samples all right
so let me go through my list and make
sure I say everything to you when we
have a large sample size greater than 30
we use a normal distribution for our
hypothesis testing okay and this is
really because of what I just said
before because the T distribution
approaches a normal distribution when
you have large number of samples greater
than 30 is our cuto off that we use now
as far as the hypothesis testing itself
the actual method is basically exactly
the same as what we've been doing before
we lock down the rejection regions we
calculate a test statistic we see where
the test statistic Falls and then we
determine if we uh reject an all
hypothesis or fail to reject so all of
that stuff is really the same the
concept is really the same but there is
one difference and that is because of
the fact that we're using a a u normal
distribution the test
statistic is as follows you you remember
before when we did uh t distribution we
had T was equal to some stuff well now
we're doing a normal distribution we use
the variable Z because that's the
variable that we use the zcore is a
variable we use with a normal
distribution but in fact this should
look very familiar to it's it's uh xar
minus mu over sample standard deviation
square root of n in fact this is
basically exactly the same thing as
before only thing I've changed is
instead of calling it t we call it Z so
really U that's why we introduced the T
distribution stuff first when we talk
about hypothesis testing because I can
get over all of these details of what it
really is and you're kind of comfortable
using it and then when we get to large
samples I just tell you hey we're doing
exactly the same thing except you're
using a normal distribution so the chart
you're going to use in your book will be
normal distribution and the test
statistic is no longer a t test
statistic because we're not using a t
distribution it's Z because that's what
we use when we talk about normal
distributions but on the right hand side
is basically what we've been doing
before um so this all this stuff on the
Right comes from the sample data that
you collect just like it did before all
right uh now here's one fundamental
difference between what we've done
before and what we've done now if you
remember I told you okay write down your
information write down a copy of or a
little sketch of the curve and then
write down your rejection regions so you
figure out do you have a left tail test
do you have a right tail test or do you
have a two tail test and you shade the
Tails and then you find those values of
T and those values of T are basically
the hurdles that you have to jump over
to get to statistical significance to
see if you reject or not I'm recapping
what we've done basically before that's
what we've done in the past and for
every one of those problems we had to go
into the T distribution and figure out
what those values of T were that Define
our rejection regions T sub alpha or
negative T sub alpha or if it's a
two-tail distribution T Alpha over two
we've done that many many many times in
all the problems we've done but you know
the reason that we had to find those uh
those values of T Alpha for every single
problem in the previous ones is because
that t distribution it totally changes
shape depending on the degrees of
freedom that you have if I have a 15
degree of Freedom problem which means 16
samples or if I have another problem
that's 22 degrees of freedom which means
you know 23 samples right then those two
are going to look Bell shape but they're
going the shape of the Curve will be
different one will be steeper than the
other because of the changing nature of
the T distribution depending on the
number of samples I have to go find
those rejection regions for every
problem that I do because the T
distribution itself the shape of it
changes for every problem depending on
the number of samples I have but now
we're into the realm of large samples
it's always going to be greater than 30
samples and I'm always going to use a
normal distribution a normal
distribution is much simpler because it
never changes the shape of the normal
distribution never ever ever changes
it's normal it's constant it doesn't
change doesn't matter if I have 32
samples or 49 samples or 110 samples I'm
still using that normal distribution for
all of these problems so these problems
actually get a little bit simpler I
don't have to go find these rejection
regions for every single problem because
that distribution doesn't change for
every problem I can kind of find them
ahead of time and I can write them down
for you so let me go and do that I'm
going to write a littleit bit of
information here and then after I get in
on the board I'll explain the
significance of what we're doing all
right so here is a column of C that's
basically a level of confidence here's
Alpha and here's one
tail
test and
two tail
test okay so here's a column here's a
column here's a column here's a column
so let's take some common values what
would we do if we wanted a 90%
confidence interval 0.90 would be C
Alpha would be 1 minus this 0.10 now
let's pretend that we were actually
going to do this full-blown we're
actually going to go find do the
hypothesis test for a large sample size
uh mean type of problem uh here so what
we would do is we'd go it's exactly the
same as before Alpha is point uh point
one so let me go down the skip down here
for a second and just sketch something
for you let's say this is the normal
distribution and let's say it was a
right tail
test okay a one tail test let's say to
the right so what I do is I take my
Alpha just like we've always been doing
in the past and I say hey this is going
to be Alpha which is
0.10 okay now this is no longer a t
distribution it's a z distribution so
there's going to be a value of Z down
here that's going to correspond at this
point such that the area to the right is
Alpha it's exactly the same as what we
had for T but now it's a normal
distribution which doesn't change the
shape of that thing doesn't change
depending on the degrees of freedom or
anything that's only something that the
T distribution you have to do deal with
so for the T distribution I had to go
and look this up for the degree of
freedom and find this value of t for
every problem to Define my rejection
regions because the shape of this curve
changes for every problem where I have a
different number of samples but here if
I really wanted to I could go look in my
table and I could go figure out the
value of Z that gives me an area of 0.1
to the right and it turns out that for a
one tail test this value of Z is going
to be
1.28 if you go look in the chart the
value of 1.28 if you do the calculations
will give you an area of 0.1 to the
right okay that is never going to change
so if you're doing a one tail test at a
90% confidence or 0.1 significance it's
the same exact thing the value of Z that
you use down here for your rejection
region is just positive 1.28 it always
is this doesn't change whether I not I
have 35 samples or 42 samples or 69
samples I don't have any degrees of
freedom or anything like that in this
problem so this value of Z that gives me
this area never ever changes that's why
I'm going to tabulate this for you okay
what if I have 0.95 that's 95% so Alpha
is 0.05 the one tail test is
1.645 that's a z value uh zv value of
1.645 here that's going to give me an
area of 0.05 to the right and it doesn't
change depending on uh degrees of
freedom or anything like that some other
common Valu
0.98 would give me 0.02 for Alpha
2.05 and then 0.99 which is 99%
confidence 0.01 level of significance
2.33 okay now this is for a one tail
test and I'm I'm writing these as
positive values of Z so I'm kind of
implying when I write it as a positive
value that it's a right-and test because
the just like just like the T
distribution this is centered at zer row
these are positive values of Z these are
negative values of Z so if I'm looking
at a right tail test where Alpha's 0.1
then I know that Z this boundary here is
1.28 okay but if I were doing a left
tail test if I were doing a left tail
test I know that my rejection region has
to have a negative value of Z but
everything is a mirror image just like
it is for T distribution so I going to
put a negative sign in front of this and
I would put a negative sign in front of
this negative sign in front of this
negative sign in front of this if I were
doing a tail test so you have to know
what you're doing and the little minutia
of details okay now I haven't done
anything with this column yet because I
wanted to explain what I'm basically
doing here if I have a two-tail test
don't forget what that is a two-tail
test is basically when you have
something like this and you have an area
over on one side and an area over on the
other side so you have two values of Z
okay but these two values of Z that you
have they're always equal and opposite
if this is 1.5 this is always negative
1.5 and what you're doing is is you're
finding the values of Z to give you the
alpha over two over here and Alpha over
2 over here so this is Alpha over 2 and
this is Alpha over 2 so you can do the
exact same thing if you grabb the normal
distribution and you found different
Alpha over tws and put them over here
you'd find a value of Z and you would
have a corresponding negative value of Z
it turns out when you do all that for
90% confidence the two tail tests is
plusus
1.645 for 95% confidence you got plus -
1. 96 for 98% confidence you've got plus
-
2.33 and for 99% confidence you've got
plus minus
2575 and that's basically the table so
um we're not going to do a problem in
this section but I want to un for you to
understand where this comes from because
a lot of books will give this to you and
a lot of students look at this and
they're just like well this is so
different than what we've done before
well I'm I explained all the details in
the T distribution and we've done before
how we arrive at these boundaries and
before you've learned from doing enough
problems that what you figure out is you
put your Alpha here your level of
significance here and you have to find
this value of T well the only reason you
had to find that value of t for every
single problem is because the shape of
the distribution changes depending on
the degrees of freedom which is the
number of samples so if I run the same
problem twice with different number of
samples I'm going to get uh I do the
have to problem do the problem twice
because the uh cut off regions will be
different if I have different number of
of samples because the shape of that
curve changes but with the normal
distribution when we get over 30 samples
we don't even have any degrees of
freedom nothing changes so what we do is
we I've done this work for you or your
book has done this work for you if you
do 90% confidence the and it's a right
tail test it would be Z is equal to 1.28
if it's a left tail test it's negative
1.28 because it'll be the mirror image
if it's a two-tail test half of the
areas in each tail so it's plus- 1. 645
you could get all of these numbers
yourself if you crack open your book to
the back to the normal distribution and
you scam through there and figure out
the areas uh and then look for the Z
values that correspond to the areas that
we're talking about the areas being
Alpha here you could get all these
yourself but you have to be careful
because remember the T distribution is
defined for area to the right of T but
the normal distribution is always
tabulated for the area left of Z so that
is just a little bit weird you kind of
have to just kind of remember that the Z
distribution the normal distribution is
the very first one you learned you get
the area to the left that's what we've
done in way in the past the T
distribution is always area to the right
so when you're if you had to do it all
manually you would have to take that
into consideration when you found find
these numbers but in reality you don't
have to find these numbers because 90 95
98 and
99% level of confidence are the most
common that you'll see certainly for all
of the problems we'll do here that's
what we'll use
so really all you have to do is when you
get to a problem and it says you're
doing 98% level of confidence or Alpha
02 you just know that the Z cut off is
2.05 if it's a right tail test if it's a
left tail test it's a negative 2.05 and
you don't have to fret about finding
those values you just use them that
defines your rejection region and then
you just slide right on into finding
your test statistic to see where it
falls so I think it's enough talking
here I think what we need to do is just
I'm going leave this on the board I'll
leave this um I'll leave this um this
chart on the board we'll solve the
problems over on the right hand side and
so as you work the problems you can
refer back to it and you'll see what I'm
talking about so F me on to the next
lesson we'll we'll work our first
problem with hypothesis testing when we
have a large number of samples testing
hypotheses involving u
means