3003PSY Mini--transcript (2)-1.pdf

SPEAKER Welcome back to the Mini Lecter syriza 33 ps y I'm Dr Madeleine Roxton. And in this mini lecture we will look a non paramedic test, starting with Humans Road. That's where in this course, all the tests that we've looked at have involved one estimating populated parameters to assumptions abou...

SPEAKER Welcome back to the Mini Lecter syriza 33 ps y I'm Dr Madeleine Roxton. And in this mini lecture we will look a non paramedic test, starting with Humans Road. That's where in this course, all the tests that we've looked at have involved one estimating populated parameters to assumptions about the shape of the data and three assumptions about the scaling of the variables. These assumptions are all related to Parametric tests. The test that we have looked at correlation or aggression are both Parametric tests, as are the T tests and an overflow wonder below three in 2000 PS y This week and next. We look at non Parametric tests, but first, let's look at the characteristics of Parametric tests first, as speed discussed previously, we used our sample data to make inferences about the population parameters. Hence the name paramedic tests. Paramedic test involved the estimation of a least one population parameter. For example. In the T test, you use the sample variance to get an estimate of the population variance. We discussed previously that we use the sample correlation to make an estimation of the population correlation. Non Parametric test did not estimate population parameters. And again, this is why they referred to as non Parametric tests. Second, Parametric tests also make assumptions about the distribution of the population from which the data were randomly sampled. All the tests we've looked at so far have all assume that the data is only distributed, For example, the teachers. The centre that sampling error is tact is distributed normally around the mean. In fact, we stand a good deal of time testing assumptions in their aggression. Regression analysis Recall how we looked at the distribution of the residuals. Non Parametric tests, on the other hand, do not make a prairie assumptions about the specific shape of the distribution and has that also known as distribution. Free tests and third paramedic test. Make assumptions about the scaly or level of the variable Parametric test assumed. The outcome or criterion Variable is measured at the interval or ratio level by very and multiple regression. Also soon that predict the variables are also measured at the interval aggression level. Although we have used the special case of dichotomous variables in week six, not here. Imagine Test do not have this assumption and actually content. We look at the case of testing predictor and outcome variables that are both categorical. So we have discovered Parametric test requires to duel over assumptions checking. They may take more to do the chicken than the actual analysis, which you may have already found. Non paramedic test did not require assumptions off normality or homage in 80 of variances. Severely skewed data can be analysed. Non Parametric stats. Non paramedic tests are ideal for analysing data from small samples. Small samples are often scared and can't be rescued by the central limit serum, not barometric tester. Also, generally easy to calculate and require less computation. This made the very popular before computer programmers made that data analysis to his friendly, none paramedic test use ranks that reduce the effect of extreme lies. Having said all this, you were to use the multiple regression in your assignment. Perform all the required tests from the previous topic. There is no equivalent Parametric Testa. Multiple regression. So what is ranking? You might ask. Ranking of all's order involves the ordering of set of schools from a small still largest. The smallest value is given the rank of one, the second smallest, the rank of to the 50th the rank of 50. I could rent student in this class by the order, but they log onto the synchronous lectures or by performance on the mid trial Mr Exam. So was ranking important, Reckon, provide the standard distribution of schools with standard characteristics. Take this example. They are six people. How excited they were to learn stats on a measure that gave me a mean score that range from 1 to 10. I may have five student to a less than thrilled but one who is very excited. The six person would be an outlaw, and we have discussed the options previously on how we can deal without lives. Another approach. They would be used. Ranking here. I'll get the person with the lowest score. Rancour won the second lowest of rancour to and so on until the high score is a given a rank of six when they wrecked, Then there is an even distance between them. The lowest score is still in. That was Frank, and the high score is still the highest rank. We can then use this wreck data, so let's let's look at Spielman's rope. You actually cover this in one double 03 PS y, but it's worth a refresher for those of you and my other course. 2016 Pierce Why you heard of Charles Spearmon in the context of the lexical hypothesis and randy cartels took so many are personality traits. This is the same Spearmon. He was based in London at University College at the same time as Cold Pearson. It was actually quite considerable animals of animosity between the two. In part they disagreed on the best approach to calculate and correlations Pearson, As you learned you said schools. His correlation. Pearson Dar is based on the assumption that the data interval a ratio that both variables are normal, distributed that there is a linear relationship between both variables and that there is a buy very normal distribution. That's a lot of assumptions here. Spearman, on the other hand, advocated the use of Frank's. You may know the Spearmint also used the Greek letter wrote to do know his correlation recall that role is also the symbol for the correlation the population. I agree that there can be a confusing to make it easier. We used the usual are with the subscript yes to indicate a spearmint correlation so Spirits wrote, is calculated using Pearsons are formulated for murder. The key difference is the data is ranked Spinetta Rohit's hand. If first the data it naturally occurred naturally falls and ranks second, if you have extreme scores an example. And third, if you have a mon atomic relationship between your variables, what is a man? A tight relationship? You may I ask. This is a non atomic relationship in this graph. As the exit next scores increases, the wife score increases or decreases, but not necessarily at a constant rate. This is nonlinear. This is a representation of ashes, conformity, conformity, experiment, looking at the association between the size of a crowd and the percentage of people who will conform to some norm, recalling the original experiment that this was judging line links here is the crowd size increases. There is a rapid increase in conformity until we reach about dating site conformity, and then this price out. Quite crowd size doesn't seem to matter as this is non linear relationship, then swim in. Its approach is to convert the conformity data into Rex. There's no creates a linear relationship, and Pearson's I can be used another benefit of Spearman's road is that it can be used when you have two continuous variables, but one or both has badly skewed due to extreme scores again. In your assignment, you will use transformation. Do not use humans, wrote it will not address the questions in the assignment. As you see in this example, I have six scores between 10 and 176. The first five schools are 10, 11, 12, 13 14, and then the final one is 176. The person score 176 is obviously an outlaw, but again, I can use ranking to give them equal distance and just give them the ranks of 1326 Well, that seems a bit convoluted, and reality is quite easy to perform. The analysis. Using a nun, Parametric Correlation Command and requesting Spearman's here, I ran the correlation between the and starts exam performance. I still find a significant correlation, but with a week of magnitude, one of the traders of using non parametric tests is that they're less helpful than Ben Parametric tests. So while they require more assumptions to be met, we typically prefer parametric tests of the non paramedic test as Parametric testify More powerful. Finally, just a note about the no hypothesis in non Parametric tests. The goal of any non paramedic test is to establish overall differences between two or possibly more distributions, not to identify the differences between any particular parameters and, as a result, that no hypothesis is more general. The male simply states that are samples come from identical populations, not just populations with the same mean and so rejected. Their now means that our populations differ. We'll return to this next week to summarise paramedic test involved estimated population parameters, making assumptions about the shape of the data and assumptions about scaling of the variables non Parametric ce do not ranking is one approaches by non Parametric tests. Spearman's road can be used to calculate correlations between variables that have natural ranks that have extreme scores in the sample. If there's if there's a mano tonic relationship between the variables, well, fewer assumptions are required to admit non permit. A general is powerful, and parametric tests