To ASSUME is to make an… Four assumptions for t-test hypothesis testing:
When do I do the what now? If all 4 assumptions are met: If the samples are not independent: If the variances (std. dev.) are not equal: If the data is not normal or has small sample size:
When to pool, when to not-pool Both tests are run by Minitab as “2-sample t-test” For pooled test check box – “Assume Equal Variances” For non-pooled, do not check box
Assessing Equal Variances… Often not recommended: Although pooled t-test is moderately robust to unequal variances, F test is extremely non-robust to such inequalities Pooled t-test will allow you to run an accurate test with some degree of unequal variance F-test is much more specific than pooled-t
Assessing Equal Variances… In both tests, the null hypothesis (Ho) is that the population variances under consideration (or equivalently, the population standard deviations) are equal, and the alternative hypothesis (Ha) is that the two variances are not equal.
What the F…? Use Levene's test when the data come from continuous, but not necessarily normal, distributions is less sensitive than the F-test, so use the F-test when your data are normal or nearly normal
When the F…? Ho: σ1 = σ2 Ha: σ1 ≠ σ2 High p-values (above α-level) Fail to Reject Null - indicate no statistically significant difference between the variances (equality or homogeneity of variances) Low p-values (below α-level) Reject Null - indicate a difference between the variances (inequality of variances)
How the F…? STAT – Basic Statistics – 2-Variances Enter columns of data as before Under “Options” can modify α-level of test (but why would you do that) Note that by default, MINITAB gives you the results of both the F-test and Levene’s Must decide a priori which test you plan to utilize
Significance Level The probability of making a TYPE I Error (rejection of a true null hypothesis) is called the significance level (α) of a hypothesis test TYPE II Error Probability (β) – nonrejection of a false null hypothesis For a fixed sample size, the smaller we specify the significance level (α), the larger will be the probability (β), of not rejecting a false hypothesis
I have the POWER!!! The power of a hypothesis test is the probability of not making a TYPE II error (rejecting a false null hypothesis) t evidence to support the alternative hypothesis POWER = 1 - β Produce a power curve
We need more POWER!!! For a fixed significance level, increasing the sample size increases the power Therefore, you can run a test to determine if your sample size HAS THE POWER!!! By using a sufficiently large sample size, we can obtain a hypothesis test with as much power as we want
Increasing the power of the test There are four factors that can increase the power of a two-sample t-test: 1.Larger effect size (difference) - The greater the real difference between m for the two populations, the more likely it is that the sample means will also be different. 2.Higher α-level (the level of significance) - If you choose a higher value for α, you increase the probability of rejecting the null hypothesis, and thus the power of the test. (However, you also increase your chance of type I error.) 3. Less variability - When the standard deviation is smaller, smaller differences can be detected. 4. Larger sample sizes - The more observations there are in your samples, the more confident you can be that the sample means represent m for the two populations. Thus, the test will be more sensitive to smaller differences.
Sample size Increasing the size of your samples increases the power of your test You want enough observations in your samples to achieve adequate power, but not so many that you waste time and money on unnecessary sampling
When to pair, when to not-pair Test is run by Minitab directly as “paired t-test” Used when there is a natural pairing of the members of two populations Each pair consists of a member from one population and that members corresponding member in the other population Use difference between the two sample means
When to pair, when to not-pair Paired t-test assumptions: 1. Random Sample 2. Paired difference normally distributed; large n 3. Outliers can confound results Tests whether the difference in the pairs is significantly different from zero
Paired Test - Example For Example… If you are testing the effects of some experimental treatment upon a population e.g. – effect of new diet upon a single sample of fish However… Paired test must have equal sample sizes
When to parametric… Nonparametric procedures Statistical procedures that require very few assumptions about the underlying population. They are often used when the data are not from a normal population.
Non-Parametric Non-parametric t-test (Mann-Whitney): Tests whether the difference in the pairs is significantly different from zero Non-parametric test are used heavily in some disciplines – although not typically in the natural sciences – often the “last resort” when data is not collected correctly, low “power”
Nonparametric tests: Less powerful than parametric tests. Thus, you are less likely to reject the null hypothesis when it is false. Often require you to modify the hypotheses. For example, most nonparametric tests concerning the population center are tests about the median rather than the mean. The test does not answer the same question as the corresponding parametric procedure. When a choice exists and you are reasonably certain that the assumptions for the parametric procedure are satisfied, then use the parametric procedure. Drawbacks of Nonparametric Tests