‭Unit 5, Inferential Statistics‬

‭Topic 19 (W17D3 KC)‬

‭●‬ A ‭ confidence interval communicates how accurate a point estimate is likely to be, based on‬
‭standard deviation‬
‭●‬ ‭A confidence interval starts with the point estimate and then adds or subtracts a‬‭margin of‬
‭error‬
* ‭σ‬
‭●‬ ‭𝑥‬± ‭𝑧‬
‭𝑛‬
‭○‬ ‭When the population standard deviation is unknown, the sample standard deviation‬
‭σ‬
‭can be used as a close estimate, where the standard error is:‬‭𝑠‬ ≈
‭𝑛‬
‭‬ T
● ‭ he more confident we wish the be, the larger (and less useful) our interval will be‬
‭●‬ ‭The more information we have about a population (larger sample size), the less new‬
‭information (additional data) affects the confidence interval‬
‭●‬ ‭A 95% confidence interval means that we can be 95% sure that the population‬
‭mean lies within our interval‬
‭●‬ ‭Confidence intervals estimate the value of a population mean; they do not estimate the value‬
‭of a sample statistic or of an individual observation‬
‭●‬ T ‭ he whole point of a confidence interval is to estimate the unknown value of‬‭μ‬‭based on the‬
‭observed value‬‭of x̄‬
‭●‬ ‭t-distribution‬
‭○‬ ‭Characterized by‬‭degrees of freedom‬
‭○‬ ‭Mound-shaped, centered at 0, wider‬
‭than normal distributions‬
‭○‬ ‭Degrees of freedom = sample size - 1‬
‭■‬ ‭𝑑𝑓‬ = ‭𝑛‬ − ‭1‬
‭●‬ ‭For an x% confidence interval, the critical‬
‭value t* is the value such that x% of the area‬
‭under the curve is between -t* and t*‬
‭○‬ ‭Look up this value in a‬‭t-table‬
‭●‬ ‭Confidence-interval for a population mean‬
‭(t-interval)‬
* ‭𝑠‬
‭○‬ ‭𝑥‬± ‭𝑡‬
‭𝑛‬
‭■‬ W
‭ here t* is the appropriate‬
‭critical value from the‬
‭t-distribution with n-1 degrees‬
‭of freedom for the desired‬
‭confidence level‬
‭ ‬ ‭Conditions to use a t-interval:‬
●
‭1.‬ ‭The sample was derived from the population via SRS‬
‭2.‬ ‭Either‬‭the sample size is large (‬‭𝑛‬ ≥ ‭30‬‭)‬‭or‬‭the population‬‭is normally distributed‬

‭well class of 2028 hf ig‬ ‭Page‬‭21‬

‭Study guide courtesy of Catelyn Dao, Patrick Du, Shreyas Jain‬
‭Topic 20 (W18D2 KC)‬
‭Test of Significance‬
‭1.‬ ‭Give a description of the parameter, being sure to identify the type of number (e.g.., a‬
‭mean or a proportion), the variable, and the population‬
‭a.‬ ‭Example: The parameter of interest is the average time in seconds that a student‬
‭estimated elapsed while the clip of "ABC" was played. Using μ to represent it‬
‭2.‬ ‭State competing claims about the parameter of interest. The null hypothesis states the‬
‭parameter is equal to a specific value‬
‭a.‬ ‭H‬‭0‬‭: parameter = hypothesized value (statement of no‬‭effect)‬
‭b.‬ ‭Example: H‬‭0‭:‬ μ = 10‬
‭The alternative hypothesis states what the researchers suspect or hope to be true about‬
‭the parameter. The form of the alternative hypothesis is determined by the research‬
‭question before the samples are collected‬
‭a.‬ ‭H‬‭a‭:‬ parameter < hypothesized value‬
‭b.‬ ‭H‬‭a‭:‬ parameter > hypothesized value‬
‭c.‬ ‭H‬‭a‭:‬ parameter ≠ hypothesized value‬
‭3.‬ ‭Specify the behavior of the sampling distribution under the null hypothesis. Typically‬
‭involving checking some technical conditions that need to be met, such as the Central‬
‭Limit Theorem. The initial two important conditions to check are a simple random‬
‭sample (SRS) and the sample size/normality of the distribution. Conditions will be‬
‭checked assuming the null hypothesis is true‬
‭a.‬ ‭Conditions for CLT:‬
‭i.‬ ‭Simple random sample (assumed implemented if unmentioned)‬
‭ii.‬ ‭n > 30 (If false, check if the distribution is normally distributed. If not‬
‭normally distribution, proceed with caution)‬
‭iii.‬ ‭Independent where n < 10% of overall population size (assume that it is‬
‭independent but make sure to state this in conditions)‬
‭Draw a well-labeled sketch with a mean of the hypothesized value and a standard‬
‭𝑠‬
‭deviation of‬ ‭√‬‭𝑛‬ ‭, where s is the standard deviation‬‭of the sample. Also mark the‬
‭observed mean value of your sample‬
‭4.‬ C
‭ alculate a test statistic to measure the discrepancy between the observed statistic and‬
‭the hypothesized value of the parameter. If the discrepancy is large, we have evidence‬
‭against a null hypothesis‬
‭a.‬ ‭The test statistic can be found by seeing how many standard deviations the‬
‭sample mean is away from the population mean‬
‭𝑥‬−‭μ‬
‭b.‬ ‭t =‬ 𝑠‭ ‬‭/√‬‭𝑛‬
‭5.‬ C
‭ alculate the p-value, which is the probability, assuming the null hypothesis to be true,‬
‭of obtaining a test statistic at least as extreme as the one actually observed. Extreme‬
‭meaning “in the direction of the alternative hypothesis.”‬
‭a.‬ ‭An alternative hypothesis of “not equal” means that you need the probability in‬
‭both tails‬

‭well class of 2028 hf ig‬ ‭Page‬‭22‬

‭Study guide courtesy of Catelyn Dao, Patrick Du, Shreyas Jain‬
 ‭b.‬ A ‭ positive t-statistic indicates finding the probability above it in the t-distribution‬
‭with the degrees of freedom. You then multiply that by 2 in order to calculate the‬
‭probability in both tails (done with a t-table or a calculator)‬
‭6.‬ ‭Summarize your conclusion in context. State a test decision or a comment evaluating the‬
‭strength of evidence against the null hypothesis where the test decision needs to be‬
‭made.‬
‭a.‬ ‭If the p-value is small, reject the null hypothesis‬
‭i.‬ ‭p-value below 0.05 but above‬
‭0.01 constitutes reasonably‬
‭strong evidence against the‬
‭null hypothesis‬
‭ii.‬ ‭p-value below 0.01 constitutes‬
‭very strong evidence against‬
‭the null hypothesis‬
‭b.‬ ‭If the p-value is high, do not reject the‬
‭null hypothesis‬
‭i.‬ ‭p-value above .10 constitutes little to no evidence against the null‬
‭hypothesis‬
‭ii.‬ ‭p-value below 0.10 but above 0.05 constitutes moderately strong evidence‬
‭against the null hypothesis‬
‭In some studies, the researcher can decide in advance how small the p-value‬
‭needs to be to support a null hypothesis. This cutoff is called a significance level,‬
‭denoted by 𝞪. A smaller significance level indicates stricter standards. If a‬
‭researcher specifies a level of significance in advance, you have to say you fail or‬
‭reject the hypothesis at a certain level.‬
‭Another common expression is to say that the data are statistically significant if it‬
‭is unlikely to have occurred by chance or sampling variability alone.‬
‭Proceed with responding to the research question and whether you have evidence for the‬
‭alternative hypothesis or not (restate final conclusions towards research question)‬
‭ ‬ ‭Hypotheses are always about parameters, not statistics‬
●
‭●‬ ‭The alternative hypothesis should be composed before data is collected (unbiased)‬
‭𝑠‬
‭●‬ ‭The denominator of a test statistic is the standard error denoted by‬ ‭√‬‭𝑛‬
‭●‬ W ‭ hen calculating the p-value for a two-sided alternative, include the total area in both‬
‭tails of the t-distribution beyond the value of the test statistic. You can calculate this total‬
‭area by doubling the area in the right tail.‬
‭●‬ ‭A low p-value indicates strong evidence against a null hypothesis‬
‭●‬ ‭You cannot generalize to a larger population if the sample was not random, in which case‬
‭you can generalize to smaller, more representative populations‬

‭well class of 2028 hf ig‬ ‭Page‬‭23‬

‭Study guide courtesy of Catelyn Dao, Patrick Du, Shreyas Jain‬
‭Topic 22 (W19D1 KC)‬
‭tl; dr: it’s basically the same as before, just with two variables‬
‭main differences (by step):‬
‭1.‬ ‭state two parameters, one for each sample group‬
‭2.‬ ‭null hypothesis is always µ‬‭1‬ ‭= µ‬‭2‭;‬ choices for alternative‬‭hypothesis are‬
‭a.‬ ‭µ‭1‬‬‭> µ‬‭2‬
‭b.‬ ‭µ‭1‬‬ ‭< µ‬‭2‬
‭c.‬ ‭µ‭1‬‬ ‭≠ µ‬‭2‬
‭3.‬ ‭for the independence and randomness conditions, you now should either have that each‬
‭sample was taken using an SRS or the two sample groups came from one large group‬
‭randomly assigned to two treatment groups‬
‭𝑥‭1‬ ‬−‭𝑥‭2‬ ‬
‭4.‬ ‭different formula:‬‭𝑡‬ = ‭2‬ ‭2‬
‭𝑠‬ ‭1‬ ‭𝑠‬ ‭2‬
‭𝑛‭1‬ ‬
+ ‭𝑛‭2‬ ‬

‭5.‬ b ‭ asically the same, just make sure to take the degrees of freedom of the lower of n‬‭1‬ ‭and n‬‭2‬
‭for your t-test‬
‭6.‬ ‭again, basically the same, just make sure your context is correct for the scenario‬

‭confidence interval (measured for µ‬‭1‬ ‭- µ‬‭2‬‭) :‬

‭2‬ ‭2‬
* ‭𝑠‬ ‭1‬ ‭𝑠‬ ‭2‬
‭1.‬ ‭different formula:‬(‭𝑥‬‭1‬ − ‭𝑥‭2‬ ‬) ± ‭𝑡‬ ‭𝑑𝑡‬ ‭𝑛‬‭1‬
+ ‭𝑛‭2‬ ‬

‭a.‬ ‭make sure to get the t* value corresponding to the lesser degrees of freedom‬

‭Topic 23 (W19D2 PKC)‬

🥳 Wooo last topic of RS1!‬
‭

I‭ f the sampling or experimental design is‬‭paired‬‭,‬‭then use‬‭paired t-procedures‬‭. If the samples‬

‭are drawn independently for the two groups, or if randomization is used to assign subjects to‬
‭separate treatment groups, use two-sample t-procedures.‬

‭ ata are collected with a paired design when there is a link between each observation‬
D
‭in one group with a specific observation in another‬

‭A‬‭paired t-procedure‬‭applies one sample t-procedures‬‭to the differences within a pair‬

‭●‬ ‭𝐻‬‭0:‬ ‭‭μ

‬ ‬‭𝑑‬ = ‭0‬

‭𝑥‬‭𝑑‬
‭𝑡‬ =
‭𝑆‬‭𝑑‭/‬ ‬ ‭𝑛‬

‭well class of 2028 hf ig‬ ‭Page‬‭24‬

‭Study guide courtesy of Catelyn Dao, Patrick Du, Shreyas Jain‬
 ‭●‬ f‭ or a p-value based on a t-distribution with (n-1) degrees of freedom, where n is the number‬
‭of pairs in the sample‬

* ‭𝑆‬‭𝑑‬
‭𝑥‬‭𝑑‬ ± ‭𝑡‬ ‭​‬

‭𝑛‬

‭●‬ ‭confidence interval for the population mean difference‬‭μ‭𝑑‬ ‬

‭ he technical conditions for paired t-procedures are the same as with a one-sample t procedure‬
T
‭except observational units are paired and the data are differences.‬

‭well class of 2028 hf ig‬ ‭Page‬‭25‬

‭Study guide courtesy of Catelyn Dao, Patrick Du, Shreyas Jain‬