EGT1 Task 3

We conduct an independent sample t-test using Excel, and obtain the following output (see sheet T-TEST)

7. Why are the larger t ratios more likely to be statistically significant? The larger t ratios are more likely to be statistically significant because it takes a larger difference between the treatment and comparator with smaller variation. The larger the t ratios the more confident we are about our results. In terms of mathematics, the 99% of the observation for t distribution falls between -3 and +3. So if the t ratio lies outside these values we are more likely to get significant results.

Of the 17 patients 5 were not taking MPH immediately prior to the study while 12 were. The stop signal test had a result that showed minimal different in reaction times between the MLR MPH and IR MPH but both were superior to the placebo pill. For the Errors of omission there was slightly any difference between the MLR and IR MPH yet both were still superior when compared the the mean score of the placebo. The arithmetic test also showed minimal difference between the two medications but the medications were superior compared to the placebos. For the IOWA -C tests significant differences were seen between the IR MPH and placebo but not from the MLR

A two-tailed t-test yields a t-value of 0.99. For 13 degrees of freedom, t-critical is 2.16 for a 95% confidence level and 1.35 for an 80% confidence level. This is reflected in Table 1 below.

In phase 1 of this trial they had a pool of 66 healthy participants that are women who are experiencing postmenopausal symptoms. The method for this experiment is to have an interval of the drug. Placing the group of women into 3 different groups which consisted of a group that had 3 mg of the drug to a group with 1.5 mg to a group that is receiving a placebo. This test is also being conducted in a way that has both the researcher and participants blinded in a random double blind test. The drug is gonna be used in an oral contraceptive with a specific dosage. This method is best used in this stage because it decreases the possibility of researcher bias. According to Dan Cartwright(2013), “Test drug, and during two periods they received single doses of both the estradiol and progesterone Reference drugs, which were taken concurrently. Blood samples were collected at multiple intervals, beginning one hour prior to start of dosing and continuing to 48 hours after

Do you think our sample size is big enough to be meaningful? What effect would taking more samples have on the reliability of our data?

In order to test if vitamins cause crime, I will conduct the double-blind research study by randomly sorting my psychology class into two groups. One group will randomly take vitamin pills and the other will get placebo pills. A placebo has no effect and is a control in most experiments. The experimental group will be the ones taking the vitamin pills while the control group will have the placebo effect. With this random assignment, the researcher and the participants would not know who is in the experimental or control group. After groups are randomly chosen, research on the crime each group commits is conducted.This way, the results for vitamins causing crimes is either proven right or wrong.

Participants: To conduct this experiment fifty-two college students will be selected. A sample test will be given out before any final decisions are made on who will partake in this study. This sample test will insure that the participants’ level of intelligence are similar. Twenty-six students, men and women, will make up the experimental group, and twenty-six will make up the control group. In each group there will be thirteen men, and thirteen women, to make the twenty-six. Each group member is placed in a certain group based on his/her sample test score.

The results shown are the average of 10 different experiments ± SEM. (*) Indicates significant difference regarding to control group p ≤ 0.05 by Student “t” test.

Samples size of this study was chosen on the basis of practical considerations rather than statistical estimation. However, according to our experience, the calculated sample size will most likely allow to detect large differences (if any) in parameters (> 50%) between the treatments groups.

I calculated the means and standard deviations of the samples and put them at the bottom of each. From this the observed difference of means came out to be 0.0072. Also from this I calculated the standard error which came out to be 0.0121. I then found the T-score by dividing the difference by the standard error and got a t-score of 0.595. The calculator did the rest of the work for me. It got a 34.7 degrees of freedom and a p-value of

Size of the sample group is needed to be specified properly due to the goal is to make inferences about the population from the sample. The appropriate sample size can be resulted in terms of the reliability of the study, especially empirical study. Therefore, how many samples can satisfy the target? In addition, a smaller sample size leads to a higher error of the sampling when compared with a bigger size. Even if an increasing of sample size can diminish the error, after it goes beyond some point, an error cannot be decreased significantly (see figure 4.1). Furthermore, an optimal sample size can also save time and cost of study.

The t‑test for differences between two means has been a standard tool for more than a century, since William Gosset introduced the method in 1908 (Gosset (Student) 1908). We use the t‑test to infer from a sample a range of values for the true mean of the population from which the sample was drawn. When comparing two independent samples we use a t‑test to decide whether the averages of the two groups are probably different by testing whether the difference in means is sufficiently different from zero. Typically we conclude that if the ratio of the estimated pooled standard error to the estimated mean difference is greater than a value determined by the sample size (usually something around 2) then there is a “significant” difference – a t‑test. Alternatively, we look up the p‑value to learn what the odds are of finding such a mean and variance purely as a result of random error in sampling. If the p‑value is less than 0.05 then we conclude that the observed difference in means is sufficiently unlikely to be zero.

Table 1: This table summarizes the two-sample t-test analysis from SPSS of the estimated dry weight statistics at sites 1 and 2. At each site, there was a sample size of 20. The mean of

be normal. Due to this reasons, parametric tests could not be applied to this data.