StatQuest with Josh Starmer · YouTube · 6:41
This 180-second clip is an introductory StatQuest segment that previews confidence intervals and then spends most of its time refreshing bootstrapping. Using a sample of 12 female mouse weights on a number line, the video distinguishes the sample mean from the population mean, demonstrates one bootstrap resample with duplicates allowed, and then shows many repeated bootstrap means accumulating into a distribution. The final seconds rename the slide to "95% Confidence Intervals," signaling that the bootstrap distribution will be used next, but the clip ends before the interval itself is constructed. This 180-second segment teaches confidence intervals through a bootstrap picture on a number line. It first defines a 95% confidence interval as the interval covering 95% of the bootstrapped means, then generalizes to a wider 99% interval. Next it reframes the interval as a visual statistical test: values outside a 95% interval are said to have and to be significantly different. Two worked examples follow. In the female-mice example, the region left of 20 is highlighted as lying outside the interval, so the video concludes the true mean being below 20 is unlikely. In the two-sample example, separate confidence intervals for female and male mice do not overlap, and the video uses that visual fact to declare a statistically significant difference with . This video segment explains how to visually assess statistical significance between two sample means using 95% confidence intervals. It establishes that non-overlapping intervals indicate a significant difference (). However, it highlights a crucial caveat: if the intervals overlap, the difference might still be significant, meaning a formal t-test is required to draw a definitive conclusion.
Use the learning inspector for key ideas and moments, or open the reading tabs for the complete notes.
Generated from the video's visuals and explanation; not verbatim speech.
The clip opens with the StatQuest title and then announces the topic: confidence intervals.
Before defining anything formally, the narrator frames the lesson by saying confidence intervals are often misunderstood, and he presents bootstrapping as one calculation method that makes the idea easier to grasp. He also notes that other methods are common in practice.
The video then switches to a "Bootstrap Refresher!" slide with a blue number line labeled 15, 20, 25, 30, and 35. Twelve red dots are added to represent a sample of 12 female mouse weights.
A red vertical line marks the mean of those 12 observations. The key distinction made here is that this is only the sample mean, not the mean of every female mouse in the world. The inferential target is the larger population mean, and bootstrapping is introduced as a way to reason about plausible values for it.
The bootstrap procedure is then laid out in three visible steps. Step 1 says to randomly select 12 weights from the original sample, and the slide explicitly notes that duplicates are OK. A second number line appears below the first to show one bootstrap sample.
The animation highlights the meaning of sampling with replacement: one original observation on the far left is chosen twice, while a neighboring observation is not included at all. This shows that bootstrap samples reuse the original data rather than collecting new measurements.
Step 2 instructs the viewer to calculate the mean of that random sample. Step 3 says to repeat steps 1 and 2 until many means have been calculated, with the slide giving >10,000 as the suggested scale.
A third number line fills with many red vertical lines, each representing one bootstrap replicate mean. Visually, this dense cluster is the bootstrap distribution of the sample mean generated from repeated resampling of the original 12 observations.
The narrator closes the refresher by saying that this is all there is to bootstrapping, and the slide title changes to "95% Confidence Intervals" while the bootstrap distribution remains on screen.
The clip ends just as the lesson pivots from constructing the bootstrap distribution to using it for confidence intervals; the actual rule for extracting a 95% interval is not shown within this excerpt.
The clip opens on a number line labeled “95% Confidence Intervals.” Red vertical marks represent bootstrapped means, and a black horizontal bar is placed beneath the central cluster. The narration defines the bar as an interval that covers 95% of those means, making the confidence interval concrete as a visual summary of the bootstrap distribution.
The same idea is then extended to a 99% confidence interval. A longer black hint line appears below the prompt, and the speaker states that the 99% interval is wider because it must cover 99% of the bootstrapped means rather than only 95%.
The next slide asks why confidence intervals are useful. The answer given is that they act as statistical tests performed visually. Since the 95% interval contains 95% of the means, the video infers that anything outside it occurs with probability less than 0.05, and therefore such values are called significantly different.
A female-mice example begins. The number line now marks the sample mean near 25 and explains that this sample mean estimates the true mean for all female mice. This sets up the target of the upcoming test: not the observed sample alone, but the population mean behind it.
The test question becomes whether the true mean is less than 20. The confidence interval is drawn again, and a green oval highlights the entire region to the left of 20. Because that highlighted region lies outside the 95% interval, the video concludes that the probability of the true mean being in that region is less than 0.05, so the difference is statistically significant.
Finally, the lesson compares two samples. The upper line shows female mice with red bootstrap marks and a black interval; the lower line shows male mice with blue bootstrap marks and its own black interval. Since the two 95% confidence intervals do not overlap, the video states that there is a statistically significant difference between the groups and that the p-value is less than 0.05.
When comparing two sample means, you can often determine statistical significance just by looking at their 95% confidence intervals. If the intervals do not overlap, as shown with the female and male mice weights, you can confidently conclude that there is a statistically significant difference between the means, meaning the p-value is less than 0.05.
However, there is an important caveat to this visual rule. Let's shift the sample means slightly to the left so that their 95% confidence intervals now overlap. Does this overlap mean the difference is no longer significant?
Not necessarily. Even when 95% confidence intervals overlap, there is still a chance that the true means are significantly different from each other. Because the visual inspection is inconclusive in this scenario, you cannot rely solely on the graph. You must perform a formal t-test to accurately determine if the difference between the two means is statistically significant.
In summary, non-overlapping 95% confidence intervals guarantee a significant difference, but overlapping intervals require further testing with a t-test to make a definitive conclusion.
The video introduces confidence intervals as the lesson topic and states that there are many ways to calculate them. Bootstrapping is presented as one method, chosen here because the narrator finds it especially intuitive. The clip therefore begins by refreshing bootstrapping before returning to confidence intervals.
Using 12 weighed female mice as the example, the video distinguishes the mean of the observed sample from the mean of all female mice worldwide. The red vertical line on the first number line marks the sample mean only; the population mean remains the quantity to be inferred.
To bootstrap the sample, the video randomly selects 12 weights from the original 12 observations. Duplicates are allowed, and some original observations may be omitted. This is explicitly called sampling with replacement.
After forming one bootstrap sample, the next step is to calculate the mean of that random sample. This replicate mean becomes one observation in the eventual bootstrap distribution.
The video instructs the viewer to repeat resampling and mean calculation until there are many replicate means, with the slide suggesting more than 10,000. The accumulated result is a distribution of bootstrap sample means.
On the third number line, each red vertical line corresponds to one bootstrap replicate mean. Together they form the empirical bootstrap distribution used to discuss plausible values for the population mean.
At the end of the clip, the title changes to "95% Confidence Intervals" while the bootstrap distribution remains visible. This signals that the next step will use the bootstrap distribution to define an interval, but the interval-construction rule is not reached before the clip ends.
In this bootstrap-based introduction, a 95% confidence interval is defined as the interval that contains 95% of the bootstrapped means. The black bar under the red tick marks is the visual representation of that interval.
Replacing 95% coverage with 99% coverage requires a larger interval on the same bootstrap distribution. The video explicitly hints that the 99% confidence interval is wider than the 95% one.
The clip presents a practical method: put a candidate value on the same number line as the confidence interval and judge whether it lies inside or outside. Outside values are treated as unlikely at the 0.05 level.
From the premise that the interval covers 95% of the means, the video takes the complement and concludes that anything outside occurs less than 5% of the time. This is used as the basis for declaring statistical significance.
Before testing a claim about the population, the video identifies the sample mean near 25 as an estimate of the true mean for all female mice. This distinguishes the observed statistic from the unknown population parameter.
The female-mice example asks whether the true mean could be less than 20. The region is highlighted in green and shown to lie outside the 95% confidence interval, so the video concludes and calls the difference statistically significant.
When comparing female and male mice, the video uses a simple visual criterion: if the two 95% confidence intervals do not overlap, then the group difference is statistically significant and the p-value is less than 0.05.
If the 95% confidence intervals of two sample means do not overlap, it indicates a statistically significant difference between the means (). This allows for a quick visual assessment without performing complex calculations.
If the 95% confidence intervals of two sample means overlap, it does not guarantee that the difference is not statistically significant. In such cases, a formal t-test must be conducted to determine the true significance.
Explore conditions, steps and evidence. Supplementary explanations are labeled separately from content shown in the video.
The narrator says they weighed 12 female mice and later says to randomly select 12 weights from the original sample.
The first number line shows 12 red dots representing the original sample.
Original sample size: the number of female mouse weights in the initial sample.
Positive integer; here fixed at 12.
On-screen step 3 reads: "Repeat steps 1 and 2 until you have calculated a lot of means (>10,000)".
The narrator says to repeat until many means are calculated, sometimes more than 10,000.
>10,000
Suggested number of bootstrap replicate means to calculate for the bootstrap distribution.
Large positive integer count of replicate statistics.
Slide title changes to "95% Confidence Intervals".
The narrator begins, "Usually when you see a confidence interval out in the wild, it's called a 90..." before the clip ends.
The spoken sentence is cut off before the narrator completes the term; only the on-screen title clearly gives 95%.
95%
Confidence level named in the slide title introducing the next topic.
Percentage between 0% and 100% used as a confidence level.
The narrator repeatedly says “95% confidence interval.”
Slide text includes “95% Confidence Intervals” and “A 95% confidence interval is just an interval that covers 95% of the means.”
95% confidence interval
An interval covering 95% of the bootstrapped means.
Applied to bootstrap means on a number line.
The narrator asks what a 99% confidence interval is and then answers it.
Text asks “Can you guess what a 99% confidence interval is?”
99% confidence interval
An interval covering 99% of the bootstrapped means.
Used as a wider analogue of the 95% interval.
“So here we have a black bar that spans 95% of the bootstrapped means…”
A thick black horizontal bar appears under the red tick marks.
black bar
Visual representation of the confidence interval on the number line.
Shown beneath the distribution of bootstrapped means.
Narrator refers to “the bootstrapped means that we just calculated.”
Many vertical red lines are plotted along a blue number line.
red vertical ticks
Individual bootstrapped mean values for female mice.
Displayed on the horizontal axis from about 15 to 35.
A blue horizontal axis with arrows and labeled ticks 15, 20, 25, 30, 35 is visible throughout.
blue number line
Axis of measured mean values used to display bootstrap results.
Numeric scale roughly spanning 15 to 35.
Narrator says “the p-value of anything outside of the confidence interval is less than .05.”
Text includes “p-value < 0.05”.
p-value
Probability statement used to judge whether a value is unlikely relative to the confidence interval.
Compared against the threshold 0.05.
“…less than .05, and thus significantly different.”
Text states “p-value < 0.05”.
0.05
Significance threshold corresponding to the complement of a 95% interval.
Used as cutoff for statistical significance.
“The sample mean is an estimate of the true mean for all female mice.”
A red vertical marker near 25 is labeled by an arrow as the sample mean.
sample mean
Estimate of the population mean based on the observed sample.
Shown near 25 on the female-mice number line.
Narrator repeatedly contrasts the sample mean with the “true mean” of all female mice.
Text asks about the probability that the “true” mean is < 20.
true mean
The population mean being estimated by the sample and bootstrap procedure.
Discussed for all female mice and compared to candidate values such as 20.
Title slide reads "StatQuest" and then "Confidence Intervals!!!".
Narrator says today's StatQuest is all about confidence intervals and that bootstrapping is one way to calculate them.
The video introduces confidence intervals as the main subject and frames bootstrapping as one method for calculating them. It also states that other calculation methods are commonly seen outside this lesson.
Confidence intervals can be calculated in multiple ways.
Bootstrapping is presented as the easiest way for the narrator to understand them.
Slide titled "Bootstrap Refresher!" lists steps 1-3 for bootstrapping a sample of 12 weights.
Narrator explains random selection of 12 weights, calculating the mean, and repeating many times.
Red dots are selected into a second number line, then many red vertical lines appear on a third number line.
The bootstrap refresh shown in the video consists of three steps: (1) randomly select 12 weights from the original sample of 12, allowing duplicates; (2) calculate the mean of that resampled set; (3) repeat steps 1 and 2 many times, with the slide suggesting more than 10,000 replicate means. The resulting collection of replicate means forms the bootstrap distribution used to reason about plausible values of the population mean.
Start from an original sample.
Resample with replacement to the same size as the original sample.
Compute the statistic of interest for each bootstrap sample.
Repeat many times; the slide suggests >10,000 replicates.
Narrator says the sample mean is not the mean for all mice on the entire planet, just the mean of the mice sampled.
A red vertical line marks the mean on the original 12-point number line.
The video distinguishes the mean of the observed sample from the mean of the whole population. The marked red line on the first number line represents the sample mean of the 12 weighed mice, while the broader goal is to use that sample information to infer reasonable values for the worldwide mean of all female mice.
The sample mean is computed from observed data only.
It is not assumed to equal the population mean exactly.
Step 1 text explicitly says "duplicates are OK".
Narrator says this is called sampling with replacement.
One leftmost point is circled and shown twice in the bootstrap sample, while a neighboring point is absent.
Sampling with replacement means that when building a bootstrap sample from the original data, the same observation may be chosen more than once and some original observations may be omitted. The video illustrates this by showing one original point duplicated in the bootstrap sample and another nearby point not included.
Each draw is made from the original sample.
Selected items are effectively returned so they can be selected again.
Bootstrap sample size matches the original sample size in this example.
“A 95% confidence interval is just an interval that covers 95% of the means.”
Same sentence appears on screen.
Black bar spans the central cluster of red bootstrap ticks.
The video defines a 95% confidence interval operationally as the interval that contains 95% of the bootstrapped means. The black bar under the red tick marks is presented as that interval.
The interval is built from previously calculated bootstrapped means.
The displayed example concerns sample means on a number line.
“It’s just an interval that covers 99% of the means that you calculated when you bootstrapped the sample.”
Prompt asks what a 99% confidence interval is.
The same definition is generalized: a 99% confidence interval covers 99% of the bootstrapped means. The narrator also states that it is wider than the 95% interval.
Applies to the same bootstrap distribution of means.
Higher coverage percentage yields a wider interval.
“I think confidence intervals are useful because they are statistical tests performed visually.”
“Confidence intervals are statistical tests performed visually.”
The method introduced here is to use the confidence interval itself as a visual decision rule: values inside the interval are treated as likely, while values outside are treated as unlikely at the 0.05 level.
Uses a 95% confidence interval.
Comparison is made against candidate values placed on the same number line.
“Because the interval covers 95% of the means, we know that anything outside of it occurs less than 5% of the time.”
Same wording appears on the slide.
From the claim that the interval covers 95% of the means, the video infers that any value outside the interval has probability less than 0.05. This is the basis for calling such values significantly different.
The interval must cover 95% of the relevant distribution.
The conclusion is stated for values outside the interval.
“The sample mean is an estimate of the true mean for all female mice.”
Arrow label states the same idea.
Red marker near 25 is identified as the sample mean.
Before using the confidence interval for testing, the video identifies the sample mean as an estimate of the population (“true”) mean for all female mice.
Applies to the female-mice sample shown on the number line.
“To perform that test, we draw our confidence interval… We can see that the area left of 20 … are outside of our 95% confidence interval.”
Text explains that the highlighted region is outside the 95% confidence interval and therefore has probability < 0.05.
Green oval highlights the region left of 20.
The procedure is: draw the 95% confidence interval for the bootstrapped means, locate the candidate value (here 20), and check whether the region of interest lies outside the interval. If it does, the video concludes and calls the difference statistically significant.
Requires a 95% confidence interval for the sample mean.
The candidate value or region must be comparable on the same number line.
“Because the 95% confidence intervals do not overlap, we know that there is a statistically significant difference…”
Text states non-overlap implies p-value < 0.05.
Two separate black bars under red and blue bootstrap ticks do not overlap.
For comparing female and male mice, the video uses the rule that if the two 95% confidence intervals do not overlap, then the difference between the groups is statistically significant and the p-value is less than 0.05.
Both groups must have 95% confidence intervals constructed in the same way.
The intervals must be visually disjoint on the common scale.
Because the 95% confidence intervals do not overlap, we know that there is a statistically significant difference in the weights of female and male mice.
If the confidence intervals overlap, there is still a chance that the means are significantly different from each other, so, in this case, you still have to do your t-test...
But when the confidence intervals do not overlap, then you can rest assured that there's a statistically significant difference between those two means.
If the 95% confidence intervals of two sample means do not overlap, it indicates a statistically significant difference between the means (). If they do overlap, a formal t-test must be conducted to determine if the difference is significant.
Comparing two independent sample means
Using 95% confidence intervals
Narrator says there are lots of ways to calculate confidence intervals and bootstrapping is just one of them.
Confidence intervals can be calculated in many different ways; bootstrapping is only one such method.
The statement concerns general statistical practice rather than a single dataset.
Universal in spirit: many possible methods exist.
Narrator says bootstrapping and the available data can determine what values would be reasonable for the global, worldwide mean of all female mice on the planet.
The clip does not yet show the formal rule for turning the bootstrap distribution into a numerical interval; that construction is deferred beyond 180 seconds.
From the observed sample, repeated bootstrap resampling can indicate which values are reasonable for the population mean.
An original sample is available.
The bootstrap procedure described in the video is applied repeatedly.
Existential/procedural: repeated resampling produces evidence about plausible values.
“A 95% confidence interval is just an interval that covers 95% of the means.”
Identical sentence on screen.
The statement is presented as a simplified bootstrap-based definition rather than a formal frequentist theorem.
In this bootstrap presentation, a 95% confidence interval is an interval that covers 95% of the bootstrapped means.
A set of bootstrapped means has been calculated.
The interval is chosen to contain 95% of those means.
For the displayed bootstrap distribution of means.
“Here’s a hint. It’s wider than a 95% confidence interval.”
A 99% confidence interval is wider than a 95% confidence interval.
Both intervals are constructed from the same bootstrap distribution of means.
For the same sample and bootstrap procedure.
“Because the interval covers 95% of the means, we know that anything outside of it occurs less than 5% of the time.”
“That is to say, the p-value of anything outside of the confidence interval is < 0.05…”
The video does not specify one-sided versus two-sided p-values or the exact hypothesis being tested.
If a value lies outside the 95% confidence interval, then its associated p-value is less than 0.05 and it is called significantly different.
The interval covers 95% of the relevant means.
The comparison is made against values outside that interval.
For anything outside the confidence interval.
“Thus, the p-value is less than .05. This is unlikely, and because of this, we can say there’s a statistically significant difference between the true mean and any value less than 20.”
Text says the highlighted region is outside the 95% confidence interval and therefore probability < 0.05.
The phrase “difference between the true mean and any value less than 20” is informal; the video does not state a formal null hypothesis.
For the female-mice example, values less than 20 lie outside the 95% confidence interval, so the p-value is less than 0.05 and the difference is declared statistically significant.
The 95% confidence interval for the female-mice bootstrap means is already known.
The candidate region is .
For all values less than 20 in the displayed example.
“Because the 95% confidence intervals do not overlap, we know that there is a statistically significant difference in the weights of female and male mice.”
“You know the p-value is < 0.05 just by looking at this picture!”
This is presented as a visual rule of thumb; the video does not prove it or discuss cases where overlapping intervals may still yield significance.
If the 95% confidence intervals for two samples do not overlap, then the difference between the two sample means is statistically significant with p-value < 0.05.
Each group has a 95% confidence interval constructed from bootstrapped means.
The two intervals are disjoint on the common scale.
For the two displayed samples, female mice and male mice.
Because the 95% confidence intervals do not overlap, we know that there is a statistically significant difference in the weights of female and male mice.
You know the p-value is < 0.05 just by looking at this picture!
If the 95% confidence intervals of two sample means do not overlap, then the difference between the means is statistically significant with a p-value less than 0.05.
The confidence intervals are 95% confidence intervals
The confidence intervals do not overlap
For any two sample means with non-overlapping 95% confidence intervals.
If the confidence intervals overlap, there is still a chance that the means are significantly different from each other, so, in this case, you still have to do your t-test...
If the confidence intervals overlap, there is still a chance that the means are significantly different from each other.
If the 95% confidence intervals of two sample means overlap, there is still a possibility that the means are significantly different, requiring a formal t-test to confirm.
The confidence intervals are 95% confidence intervals
The confidence intervals overlap
For any two sample means with overlapping 95% confidence intervals.
Numbered steps 1-3 appear sequentially on the slide.
First the original sample is shown, then one bootstrap sample, then many bootstrap means as vertical lines.
Narrator walks through selecting 12 weights, calculating the mean, and repeating many times.
Begin with the original sample of 12 observed female mouse weights plotted on a number line.
This is the starting data shown in the video before any resampling.
Randomly select 12 weights from the original sample, allowing duplicates, to form one bootstrap sample.
This is step 1 on the slide and is illustrated by moving red dots to a second number line.
Calculate the mean of that bootstrap sample.
This is step 2 on the slide; the video treats the mean as the statistic of interest.
Repeat the resampling-and-mean-computation process many times, with the slide suggesting more than 10,000 replicate means.
This is step 3 on the slide and is reinforced by the narration.
Collect all replicate means on a number line, producing a dense distribution of bootstrap means.
The third number line fills with many red vertical lines, visually representing the bootstrap distribution.
Repeated bootstrap resampling of the original sample produces a distribution of sample means that the video uses as the basis for discussing confidence intervals.
Narrator contrasts the 12 sampled mice with every female mouse on the planet.
A red vertical line marks the mean of the displayed sample.
Only 12 female mice were weighed, not all female mice in the world.
The narrator explicitly states this limitation of the data.
Therefore the mean computed from those 12 values describes only the sampled mice.
The video says the sample mean is not the mean for all mice on the entire planet.
The purpose of bootstrapping is to use the observed sample to assess reasonable values for the larger population mean.
The narrator directly links bootstrapping to determining plausible worldwide mean values.
The red line on the first number line should be interpreted as a sample statistic, while the inferential target is the unknown population mean.
Narrator moves from “covers 95% of the means” to “anything outside of it occurs less than 5% of the time.”
Slide text presents the same inference.
The derivation is heuristic and tied to the bootstrap framing rather than a formal sampling-theory proof.
Start from the video’s definition that the interval covers 95% of the means.
Given directly by the confidence-interval definition in the clip.
Take the complement of the covered region.
Complement rule for probabilities.
The video treats being outside the interval as evidence of statistical significance.
Stated explicitly in the narration and slide text.
Values outside the 95% confidence interval are assigned p-value < 0.05 and labeled significantly different.
Narrator draws the interval, notes the region left of 20 is outside it, and concludes .
Green oval highlights outside the black interval bar.
The clip does not define the null hypothesis formally or distinguish one-sided from two-sided tests.
Place the confidence interval on the number line.
Method stated by the narrator: “To perform that test, we draw our confidence interval…”
Mark the value being tested on the same axis.
The question on screen asks about the true mean being < 20.
Observe that the region left of 20 lies outside the interval.
Visible in the diagram and stated in narration.
Because the region is outside the 95% interval, assign probability less than 0.05.
Uses the earlier outside-tail rule.
Conclude significance against values below 20.
Explicitly stated in audio and caption.
The female-mice example concludes that values less than 20 are statistically unlikely for the true mean, with .
Narrator compares the female and male intervals and infers significance from non-overlap.
Two black bars under separate clusters of red and blue ticks do not overlap.
The rule is presented visually without proof; standard theory notes that non-overlap is sufficient but not necessary for significance.
Use the red bootstrap means and their black interval bar.
Shown and described in the clip.
Use the blue bootstrap means and their black interval bar.
Shown and described in the clip.
Check that the two intervals do not overlap.
Directly visible in the figure and stated in narration.
Conclude a statistically significant difference between the groups.
Stated as the visual decision rule in the video.
Because the female and male 95% confidence intervals are disjoint, the video declares a statistically significant difference in mouse weights with .
Narrator says, "Imagine we weighed a bunch of female mice... In this case, we weighed 12 of them."
A number line from 15 to 35 shows 12 red dots, then a bootstrap sample, then many bootstrap means.
The slide text lays out the three bootstrap steps using this 12-weight example.
Exact numeric coordinates of the 12 original dots are not labeled individually, so the precise sample mean value cannot be recovered from the clip alone.
Given a sample of 12 female mouse weights, show how bootstrapping can be used to study plausible values for the population mean.
A sample of 12 female mouse weights is plotted on a number line.
The sample mean is marked by a red vertical line.
Resampling is done from the original 12 observations only.
Construct a bootstrap distribution of sample means from the original 12 observations.
Plot the 12 observed weights on a number line and compute their sample mean.
This establishes the original sample and the statistic of interest shown in the video.
Draw a new sample of size 12 from the original 12 observations, allowing duplicates.
This is step 1 of the bootstrap procedure shown on screen.
Compute the mean of that bootstrap sample.
This is step 2 of the bootstrap procedure shown on screen.
Repeat the resampling and mean calculation many times, with the slide indicating >10,000 replicates.
This is step 3 of the bootstrap procedure shown on screen.
Display all bootstrap means as many red vertical lines on a number line.
The animation shows the accumulated bootstrap distribution after repetition.
The example yields a bootstrap distribution of replicate means built from repeated samples of size 12 drawn with replacement from the original mouse-weight data.
The video verifies the idea visually by contrasting the original 12 points, one bootstrap sample with duplication, and the final dense set of bootstrap means.
Narrator introduces the female-mice example and asks about the true mean being less than 20.
On-screen question: “What is the p-value that the ‘true’ mean of all female mice, not just our sample, is < 20?”
Red ticks, sample mean near 25, black CI bar, and green highlight left of 20.
The exact numeric endpoints of the confidence interval are not stated; only their visual positions are shown.
Determine whether the true mean weight of all female mice is plausibly less than 20, using the displayed 95% confidence interval.
Bootstrapped means for female mice are shown as red vertical ticks.
The sample mean is marked near 25.
A 95% confidence interval is drawn as a black bar.
The candidate region is values less than 20.
Assess whether is statistically significant for the true mean.
Recognize the point estimate of the population mean.
Labeled in the diagram and narration.
Use the black bar representing 95% coverage of the bootstrapped means.
Method stated in the clip.
Shade the left side of the number line with a green oval.
Shown visually and referenced in narration.
Compare the highlighted region to the interval bar.
Direct visual observation plus explicit narration.
Apply the rule that outside the 95% interval means probability less than 0.05.
Derived from the earlier coverage-to-tail argument.
The video concludes that the true mean being less than 20 is unlikely, with , so there is a statistically significant difference.
Verification is visual: the green region left of 20 is entirely outside the black 95% confidence interval bar.
Narrator says they will compare two samples, female mice and male mice, and infers significance from non-overlap.
Text states the intervals do not overlap and therefore .
Upper red female-mice interval and lower blue male-mice interval are separated.
Exact interval endpoints are not numerically given; only approximate positions on the axis are visible.
Use the displayed confidence intervals to decide whether female and male mice differ significantly in mean weight.
Female mice bootstrap means are shown in red with a black 95% CI.
Male mice bootstrap means are shown in blue with a black 95% CI.
The two intervals are drawn on parallel number lines with the same scale.
Determine whether the two groups differ significantly.
Inspect whether the black bars overlap.
This is the method introduced by the narrator.
The intervals are visually disjoint.
Directly observable in the figure and stated in audio/caption.
Conclude statistical significance from non-overlap.
Stated as the rule in the clip.
There is a statistically significant difference in the weights of female and male mice, with .
The verification is the absence of overlap between the two black confidence-interval bars on the shared scale.
Two number lines showing data points and confidence intervals for 'Female Mice' and 'Male Mice'.
Because the 95% confidence intervals do not overlap, we know that there is a statistically significant difference in the weights of female and male mice.
Determine if there is a statistically significant difference in the weights of female and male mice based on their sample data and confidence intervals.
Sample data for female mice weights
Sample data for male mice weights
95% confidence intervals for both groups
Assess statistical significance of the difference in means.
Observe the initial state where the 95% confidence intervals for female and male mice do not overlap.
Visual inspection of the diagram.
Conclude that there is a statistically significant difference () because the intervals do not overlap.
Rule for comparing two samples using confidence intervals.
Shift the means to the left so that the confidence intervals now overlap.
To illustrate the caveat of the rule.
Conclude that a formal t-test is still required because overlapping intervals do not guarantee a lack of significance.
Caveat for comparing two samples using confidence intervals.
When intervals do not overlap, the difference is significant (). When they overlap, a t-test is needed.
Visual demonstration of shifting the means to create overlap and the accompanying explanatory text.
The word "StatQuest" appears centered on a white background.
The title slide adds "Confidence Intervals!!!" below "StatQuest".
Narrator welcomes viewers to StatQuest and announces the topic.
White background
Black text "StatQuest"
Subtitle "Confidence Intervals!!!"
The logo text appears first.
The topic subtitle is added beneath the logo.
No mathematical diagram is shown yet.
The visual style remains plain black text on white.
This segment identifies the series and announces that the lesson topic is confidence intervals.
A blue horizontal number line labeled 15, 20, 25, 30, 35 appears under the heading "Bootstrap Refresher!".
Twelve red dots appear along the number line. A red vertical line is added to mark the mean.
Text appears: "Imagine we weighed a bunch of female mice..." and then "Calculate the the mean...".
The exact x-values of the 12 dots are not individually labeled.
Blue number line with tick labels 15, 20, 25, 30, 35
12 red dots
Red vertical mean marker
Explanatory text
The empty number line is introduced.
Red dots are added one by one to represent the sample.
A red vertical line marks the sample mean.
The axis scale stays fixed from 15 to 35.
The sample size remains 12 throughout this stage.
The visual encodes the observed data set and distinguishes the sample mean from the raw observations.
Step 1 text appears: "Randomly select 12 weights from the original sample (duplicates are OK)."
A second blue number line appears below the first with red dots forming a bootstrap sample. A red circle highlights a duplicated leftmost point, and arrows connect it back to the original sample.
The exact identity of every dot in the bootstrap sample is not numerically labeled.
Original sample number line
Second number line for bootstrap sample
Red dots
Red highlight circle
Black arrows
A second number line is added.
Dots are placed to form a new sample of size 12.
One duplicated point is circled and linked back to the original sample.
Both number lines use the same horizontal scale.
The bootstrap sample still contains 12 points.
The animation demonstrates sampling with replacement by showing that one original observation can appear twice while another can be omitted.
Steps 2 and 3 appear: "Calculate the mean of the random sample." and "Repeat steps 1 and 2 until you have calculated a lot of means (>10,000)".
A third number line appears and fills with many red vertical lines clustered around the center.
The exact number of displayed vertical lines is not countable reliably from the clip.
Third blue number line
Many red vertical lines
A third number line is introduced.
Red vertical lines accumulate rapidly across a central range.
The horizontal scale remains aligned with the earlier number lines.
The statistic being plotted is the mean of each bootstrap sample.
The dense set of vertical lines visually represents the distribution of bootstrap sample means produced by repeated resampling.
The slide title changes to "95% Confidence Intervals".
The previously built distribution of red vertical lines remains on the number line.
Narrator begins transitioning from bootstrapping to confidence intervals.
The clip ends before the video explains how the 95% interval is read from the distribution.
Title text "95% Confidence Intervals"
Number line with many red vertical lines
The heading changes from "Bootstrap Refresher!" to "95% Confidence Intervals".
The bootstrap distribution remains visible as the next concept is introduced.
The same number line and cluster of replicate means stay on screen.
The video signals that the bootstrap distribution just constructed will be used to define a confidence interval, but the actual interval-construction step occurs after this clip ends.
A black horizontal bar appears beneath the red tick marks after the title slide is shown.
Text explains that the interval covers 95% of the means.
Blue number line
Red vertical bootstrap ticks
Black horizontal bar
Title text “95% Confidence Intervals”
The black bar is added under the central cluster of red ticks.
Explanatory text appears below the axis.
The red ticks remain fixed in place.
The axis labels 15, 20, 25, 30, 35 remain unchanged.
The black bar visually encodes the interval containing 95% of the bootstrapped means.
A longer black line appears below the prompt about a 99% confidence interval.
Narrator says the 99% interval is wider than the 95% interval.
The exact alignment of the longer line with the upper interval is schematic rather than numerically labeled.
Existing 95% interval bar
Longer black hint line
Prompt text about 99% confidence interval
A longer black line is added beneath the question text.
The visual emphasis shifts from the 95% interval to a wider hypothetical interval.
The underlying red bootstrap ticks and axis stay the same.
The longer line illustrates that increasing coverage from 95% to 99% widens the interval.
The slide changes to “Why are confidence intervals useful?” while retaining the same number-line graphic.
New explanatory text about visual statistical tests and appears.
Title “Why are confidence intervals useful?”
Same red ticks and black CI bar
Explanatory paragraphs
Header text changes.
Additional paragraphs appear below the number line.
The bootstrap distribution graphic remains essentially unchanged.
The same interval is repurposed from a descriptive summary into a visual hypothesis-testing tool.
The slide changes to “Visual Statistical Tests”; red dots replace the earlier dense ticks, and an arrow labels the sample mean near 25.
Text states the sample mean estimates the true mean for all female mice.
The change from many red ticks to fewer red dots is a simplification of the same bootstrap display.
Blue number line
Red dots
Red vertical marker near 25
Arrow label
The display switches from dense red ticks to spaced red dots.
A marker near 25 is explicitly labeled as the sample mean.
The axis scale remains 15 to 35.
The topic remains female mice.
The clip establishes the point estimate before introducing the interval-based test.
A green oval appears around the region left of 20.
Text explains that the highlighted region is outside the 95% confidence interval.
Green oval
Number 20 tick
Black CI bar
Red bootstrap display
A translucent green oval is drawn over the left tail up to 20.
Explanatory text about probability < 0.05 appears below.
The black CI bar and axis remain in place.
The highlighted region isolates the candidate values being tested against the confidence interval.
A second number line appears below the first, labeled “Male Mice,” with blue ticks and its own black CI bar.
Text states that non-overlapping 95% confidence intervals imply significance.
Upper female-mice line with red ticks and black CI
Lower male-mice line with blue ticks and black CI
Labels “Female Mice” and “Male Mice”
A new lower number line is introduced.
Blue bootstrap ticks and a second black interval bar appear.
Conclusion text about non-overlap is added.
Both lines share the same horizontal scale.
The female-mice display remains above for direct comparison.
The visual juxtaposition supports the rule that disjoint 95% intervals indicate a significant difference between groups.
Slide titled 'Visual Statistical Tests - compare two samples'. Two number lines for 'Female Mice' (red) and 'Male Mice' (blue). Black horizontal bars represent 95% confidence intervals. The bars do not overlap.
Number line for Female Mice
Number line for Male Mice
Red data points
Blue data points
Black confidence interval bars
The confidence intervals do not overlap
Demonstrates the condition where non-overlapping 95% confidence intervals indicate a statistically significant difference between the two sample means.
Text says many people misunderstand confidence intervals because they didn't learn about bootstrapping first.
Narrator repeats that misunderstanding is common and attributes it to not learning bootstrapping first.
People often feel unsure about confidence intervals because they treat them as abstract formulas without an intuitive resampling basis.
The video argues that learning bootstrapping first makes confidence intervals easier to understand, because the interval can be grounded in a distribution built from repeated resamples of the observed data.
Narrator explicitly says the sample mean is not the mean for all mice on the entire planet.
The mean computed from 12 observed mice might be mistaken for the true mean of all female mice worldwide.
The video distinguishes the sample mean, which summarizes only the observed 12 mice, from the population mean, which is the broader quantity inferred using bootstrapping.
Step 1 says duplicates are OK.
Narrator names the process sampling with replacement.
A duplicated point is circled in the bootstrap sample.
A bootstrap sample may be wrongly assumed to require fresh or non-repeated observations.
In the video's bootstrap procedure, values are drawn from the original sample with replacement, so duplicates are allowed and some original values may be left out.
“That’s it! That’s all a confidence interval is, nothing more, nothing less.”
This is the video’s simplified pedagogical framing, not a full technical discussion of confidence-interval theory.
A confidence interval may be thought to require additional interpretation beyond covering a stated percentage of the relevant means.
The video insists that, in this bootstrap setting, a confidence interval is simply the interval covering the stated percentage of the bootstrapped means.
Hint: “It’s wider than a 95% confidence interval.”
One might guess that increasing confidence from 95% to 99% makes the interval tighter.
The video states the opposite: a 99% confidence interval is wider than a 95% confidence interval.
The clip only states that non-overlap implies significance.
This caution is not explicitly spoken in the clip; it is added as editorial context.
From the clip alone, a learner might infer that overlapping 95% confidence intervals always imply .
The video only justifies the direction “non-overlap ⇒ significant.” It does not establish the converse, and in general overlapping intervals can still correspond to a significant difference.
If the confidence intervals overlap, there is still a chance that the means are significantly different from each other, so, in this case, you still have to do your t-test...
If the confidence intervals overlap, there is still a chance that the means are significantly different from each other.
Believing that if 95% confidence intervals overlap, the difference between the means is definitely not statistically significant.
Overlapping 95% confidence intervals do not guarantee a lack of statistical significance; a formal t-test must still be performed to determine if the means are significantly different.
Narrator says bootstrapping is one way to calculate confidence intervals and then spends the clip refreshing bootstrapping before returning to confidence intervals.
The slide sequence moves from "Confidence Intervals!!!" to "Bootstrap Refresher!" and then to "95% Confidence Intervals".
The bootstrap method is presented as a tool used to build the distribution that underlies the confidence interval discussion.
Narrator first calculates the sample mean and then says to bootstrap the sample.
Step 2 of the bootstrap procedure is to calculate the mean of the random sample.
Understanding the sample mean is required before following the bootstrap procedure, because each bootstrap replicate is summarized by its mean.
Step 1 explicitly includes "duplicates are OK" within the bootstrap instructions.
Narrator defines this as sampling with replacement.
Sampling with replacement is a component of the bootstrap procedure shown in the video.
The title changes to "95% Confidence Intervals" while the bootstrap distribution remains on screen.
Narrator transitions from bootstrapping to confidence intervals.
The clip ends before the 95% interval is actually constructed or defined.
The bootstrap distribution prepared in this clip is the basis for the upcoming discussion of a 95% confidence interval, although the interval rule itself is not yet shown.
The narrator first defines the 95% interval and then asks the viewer to extend the idea to 99%.
The 99% interval is obtained by replacing 95% coverage with 99% coverage in the same bootstrap definition.
After defining the interval, the narrator asks why confidence intervals are useful and answers that they are visual statistical tests.
The coverage definition is applied as a practical rule for visually deciding whether values are likely or unlikely.
The slide links the visual-test idea to the statement that outside values have .
The visual testing method depends on the rule that being outside a 95% interval corresponds to probability less than 0.05.
The female-mice example applies the outside-the-interval rule to the region left of 20.
The one-sample example is a direct application of the outside-tail probability rule.
The narrator says, “Here’s another example… In this case, we’re going to compare two samples.”
The two-sample rule extends the single-interval visual test to comparing two separate confidence intervals.
Text states the sample mean estimates the true mean for all female mice.
Understanding the sample mean as an estimate of the true mean is needed before testing claims about the true mean with the interval.
If the confidence intervals overlap, there is still a chance that the means are significantly different from each other, so, in this case, you still have to do your t-test...
The t-test is applied when the visual method of comparing non-overlapping confidence intervals is inconclusive due to overlap.
The slide titled "Bootstrap Refresher!" lists the bootstrap steps.
Narrator explains the three-step bootstrap procedure.
Narrator says the sample mean is not the mean for all mice on the entire planet.
Step 1 says duplicates are OK.
A duplicated point is circled in the bootstrap sample.
Step 3 says to repeat until many means are calculated, >10,000.
Narrator mentions sometimes more than 10,000.
Many red vertical lines accumulate on a third number line after repeated bootstrap means are calculated.
Intro text says people misunderstand confidence intervals because they didn't learn about bootstrapping first.
Narrator repeats this explanation.
Opening definition of the 95% confidence interval.
Hint that the 99% interval is wider than the 95% interval.
“Confidence intervals are statistical tests performed visually.”
Statement that anything outside the interval occurs less than 5% of the time.
Question about the true mean of all female mice being < 20.
Explanation that non-overlapping 95% confidence intervals imply significance.
Covered · Opening branding and welcome; no mathematical content beyond identifying the series.
Covered · Introduction to confidence intervals and the claim that bootstrapping is one calculation method.
Covered · Original sample of 12 female mouse weights is plotted and the sample mean is distinguished from the population mean.
Covered · Step 1 of bootstrapping is shown: resample 12 weights with replacement from the original sample.
Covered · Steps 2 and 3 are shown: compute bootstrap means and repeat many times to build the bootstrap distribution.
Covered · Transition to the next topic, titled 95% Confidence Intervals; the actual interval construction is not reached within this clip.
Covered · Defines the 95% confidence interval as covering 95% of the bootstrapped means and shows the black interval bar.
Covered · Extends the definition to 99% intervals and states they are wider.
Covered · Explains why confidence intervals are useful and derives the outside-interval rule.
Covered · Introduces the female-mice example and identifies the sample mean as an estimate of the true mean.
Covered · Performs the one-sample visual test for values less than 20 and concludes significance.
Covered · Compares female and male mice using non-overlapping 95% confidence intervals and concludes significance.
Covered · Introduction of the rule for non-overlapping confidence intervals and its implication for p-value.
Covered · Transition to the caveat slide and animation of shifting the means to create overlap.
Covered · Explanation of the caveat regarding overlapping confidence intervals and the necessity of a t-test.
Covered · Closing remarks and end screen.
Reviewed current material from 180 seconds constructs a 95% bootstrap confidence interval from the central 95% of replicate means, compares it with a wider 99% interval, and retains the caveat that overlapping intervals require a formal test.
YouTube14:31Prerequisite in Probability and statistics: Parameter estimation → Confidence intervals.
Parameter estimation · View learning sequence