Vous êtes sur la page 1sur 4

# Kevin Fisher May 15, 2013 Psychology Statistics Yarensky 05 Final Exam: Short Answer

Final Exam: Part 1 Short Answer Questions about the Supplementary Readings in the Packet

1. In regards to hypothesis testing logic, the hypothesis follows a process. First, the independent variable has no effect is an innocent way of thinking and process a hypothesis. The next step is to collect data and prove this hypothesis wrong, this is like guilt. Then, the innocence is retained until the evidence against it seems unlikely beyond a reasonable doubt. This is rejecting the original hypothesis of innocence, leaving no alternative but the choice that the independent variable (IV) did affect the dependent variable (DV). Having a null hypothesis is that innocence that the IV does not have any effect on the DV. Generally, we want to reject this null hypothesis. The alternative hypothesis says our initial hypothesis is incorrect, that the IV did have an effect on the DV we generally want to support the alternative hypothesis. When we collect data, these are sample statistics that will be compared to the population parameters. If a sample mean is different from the population mean, this may be due to sampling error we must evaluate if the difference in data is due to the effect of the IV or just sampling error. We ask ourselves by how much wed expect the sample mean to differ from the population mean if sampling error were the only error (no alternative elements). The standard error of the mean works with the observed effect size (sample mean population mean) as the standard error provides us an estimate of the average amount of

sampling error. The statistical test takes a ratio of the actual (observed effect) size to the expected effect size. This ratio is like a z-score, but now it is being used as a statistical test and represents the computed value of z based on our actual data. The point of this testing is to prove our initial innocence wrong, which only leaves the alternative hypothesis to believe. If the null hypothesis is true, the observed effect size is only the result of sampling error. If the null hypothesis can be rejected, the observed effect size is mainly a measure of the IV. If the difference between the population mean and the sample mean is large in respect to the experiment, then we can conclude that the difference is not solely due to sampling error but due to some other element, our IV. To make this decision clearly, we set our significance level, which is a level of how confident we need to be with our results. Generally, .05 is an appropriate criterion. Now, the critical value of z table is used. If the absolute value of the test statistic is greater than the critical value of z, the null hypothesis can be rejected. There are some possible errors in interpretation of the statistical test results. A false positive, which is a Type I error, means a result that indicates a given condition has been fulfilled, when it actually has not been fulfilled. A false negative is where a test result indicates that a condition failed, while it actually was successful. A Type II error is when the null hypothesis fails to be rejected and the null hypothesis is actually false. Of course, we want to have more power in rejecting the null hypothesis -- a few factors that affect the power of our test include sample size, size of experimental effect, variability in the data, and statistical analysis issues.

2. There are many similarities between the various hypothesis tests. The tests all have the same meanings for a null hypothesis and an alternative hypothesis. The general procedure for each test is to state the hypotheses, set criterion for rejecting the null hypothesis, collect data, test the data to see if you can reject the null hypothesis, and interpret the results. This process is the same between tests. While the details change between tests, the form remains the same. In contrast, computations from test to test will differ, as well as the small details. The knowledge of population parameters is different between tests as well we do not always know the population standard deviation. Shape depends on sample size which can change, too. T-tests can be used instead of z-tests, and t-tables can be used instead of z-tables. The different types of tests include the single-sample hypothesis tests, z vs. t, two-sample hypothesis tests, and testing methods for three or more experimental conditions exists as well, which focus on analysis of variance. When a sample mean is being compared to a population mean, use a single-sample hypothesis test. Two-sample hypothesis tests are used when comparing two sample means with each other. And lastly, ANOVA tests are used for three or more experimental conditions. 3. Descriptive statistics are used to describe the characteristics of data, while inferential statistics are used to make predictions about population parameters based on knowledge of sample statistics. In regards to hypothesis testing, both z-tests and t-tests can be used. Used z-tests if the population standard deviation is known; if it is unknown, use a t-test. If you are describing the characteristics of one variable, looking for areas under the normal distribution, making a prediction of one variable based on another, looking at proportions, you are most likely dealing with a z-score. If you are testing differences between group means and finding connections, you are doing hypothesis testing. If you

have three or more samples, then there is only one method to work by, and that is analysis of variance (ANOVA). Some problems to avoid are as follows: a correlation determines strength while regression makes predictions of one variable based on another known variable. A single-sample t-test is needed in the case that the population standard deviation is not known if it is known, a single-sample z-test can be used. 4. ANOVA is used in real-life research involving three or more levels of an independent variable. This is extremely common research in the psychology field. The null hypothesis for an experiment with four experimental conditions is as follows: H0 : 1 = 2 = 3 = 4. The alternative hypothesis is H1 = Not H0. After the hypotheses are instated, like other tests, the data is collected, tested, and analyzed. In an ANOVA, the test statistic is the Fratio. It is similar to z-observed or t-observed in the other tests. It is a ratio between two numbers, one of which represents the observed effect size, and the other of which represents expected effect size. The null hypothesis is this ratio equals 1.0, or the treatment effect is the same as the experimental error. This hypothesis is rejected if the Fratio is significantly large enough that the possibility of it equaling 1.0 is smaller than some pre-assigned criteria such as 0.05. SS is the sum of squares, which is the numerator of the standard deviation/variance equation. The degrees of freedom are similar to what we are used to in an ANOVA, and MS is the mean square, which is equal to standard deviation squared, also the same as variance. F, which we have already discussed, is the final f-ratio. If the probability at the end is less than or equal to .05, we can reject the null hypothesis. If probability is bigger than .05, we have to fail to reject the null hypothesis.