Flashcards
Topic 9
Flashcards do conjunto (5)
- 1two sample t test- assume each group in the study is a sample from the respective populations
- an important assumption is that the 2 groups are independent of each other
- T-test is for comparing 2 groups and works in the same way as the one sample procedure, you take the difference between the 2 sample means and compare it to the standard error of that difference - 2different types of t test2 versions depending on whether we can assume that the 2 populations have equal variances (or SD)
- if they DO, use a POOLED t-test (slightly more powerful and generalises more easilty)
- if they do NOT- use a WELCH t test
- use levene's test to decide which version to do - 3Levene's test for 2 sample t test- if Levene's is significant there may be differences in variances, so the pooled method is not appropriate (use Welch)
- if Levene's is not significant then supports the use of the pooled tes
- use a box plot to assess the assumption for yourself
- P-value greater 0.05 is NOT significant
- then look at the 2 tailed p-value (sig) if it is greater than 0.1 then conclude there is no evidence against the null hypothesis - 4non parametric tests- a t-test is a parametric test because it explicitly estimates parameters using statistics. Since the sample mean is affected by outliers, we have seen that t-test results can be diluted by their presence
- non parametric tests aim to make inferences less directly, and are sometimes called distribution free tests because they are based on fewer assumptions (do not assume the variable has a normal distribution)
- generally more powerful and can test a wider range of alternate hypotheses
- if data is large enough that the central limit theorem applies, parametric tests are more appropriate - 5Mann-Whitney U test- replace observed values with their ranked values - this is one way nonparametric tests are protected against outliers
- the MW test is used to test the null hypothesis
- if we assume that the distributions have the same shape and spread, then evidence against the H0 will suggest that the median values are different between the groups
- the MW test works by seeing if the total ranks in one group are more that you would expect by chance, as given by the P-value
Informações
- Descrição
- HRSS3101 FlashCards sobre Topic 9, criado por Soph B em 24-06-2020.
- Criado por
- Soph B
- Visualizações
- 0
- Avaliação