 My class is on the status test, so today we are dealing with the test of significance, not going very detail about each test, but test where to apply a particular test and on what conditions or scenario. So this is one of the most scariest question we ever faced during our course graduation. Because we might not be knowing many tests and its applications, we hardly know T test and over M chi square. So when to apply and what conditions it is to be applied, it is a trickiest part of this statistical testing. So before that, I would like to suggest that please watch my videos on normal curve and types of variables before seeing this one. So if you have a clear idea about normal curve and types of variable, it is very easy to get the idea of types of testing. Otherwise it might be a roadblock for you to understand what I am talking now. So I really suggest you to watch those two videos, normal curve and types of variables. So basically, this is hardly a false lights, after false lights, we will be knowing everything and almost all the tests which are basically required for our testing and clinical studies and thesis making. So we know by statistics, it's statistics on the living organisms or human beings. So basically we do testing to create or test hypothesis. So the descriptive part of biostatistics is to create a hypothesis for we collected data of some people or 100 people. So first what we do, we do the descriptive analytics or descriptive statistics, not analytics statistics, that is to create hypothesis. So first in any article, we can see that the first two, three tables are of descriptive nature. That is the variables will be put into mean, standard deviation, median, IQR and various frequency. The questionnaire, various frequency of responses will be put. And there will be some two, three tables which are completely for this descriptive part. Then you can see the inferential part. That is the testing of the hypothesis. So there you can see many tests, correlation or regression analysis. Testing means you can see T test and over test, what will Cox and Sine try and test. So many tests. So any article or any part of research, the first part is descriptive part and second is the inferential part. Inferential part is nothing but we are trying to bring out an inference from our data. So descriptive part will be done by mean, median and standard deviation, IQR frequency most commonly and inferential test you should parametric or non-parametric test or correlation and regression analysis. So before that we need to know what is normality. Normality I am not explaining here because I have already covered a big session on normality in my previous videos. So please do check the normality video. So usually we can manually check the normality of a particular data. So suppose we get a data, we need to check whether it is normally distributed. So depending upon normality, the test which we are going to apply will be changed. So we can do manually but nowadays it's a era of softwares, we can use softwares to test whether our data is following normally normality or not. So commonly used test are Shapiro Vulcans test, Smyrna of Kolmogorov test and Anderson darling test. So you just remember this name because you might have come across these test before in all articles. One of the test will be in all articles where the statistical description has given the normality of the data was tested by Shapiro Vulcans test or Smyrna Kolmogorov test. So one sentence you might have seen. So this is just to check whether it is normally distributed or not. So it is all done by softwares. You have manual application of normality testing by taking a million standard deviation and checking the three standard deviation and to know that whether it is following 99.7% total observation. But it might take a lot of time or it is very laborious when the data is very, the data is very huge. So commonly we do, statisticians will do one of these test. Many test are there. I just put three. So this is normality. So once we check normality, we get to know that whether it is, whether our data is normally distributed or not. So if it is normally distributed, we have to follow parametric test. If it is not normally distributed, we have to follow nonparametric test. So for that, you need to know what is continuous data and what is categorical data. So this is all done on continuous data, just like height, weight, blood pressure, saliva, flow, buffering capacity. All most of the variables are continuous data. Categorical data is mostly the questionnaire response. The percentage of response, the like a scale types will come to categorical types. This category is the eye color of the blue, brown, red, the categories or the gender, male, female, the percentage of males and females in a classroom and people with diabetes and with diabetes. That proportion or categories come, it becomes categorical. But normal variables are continuous, like height, weight, age, blood pressure, so such variables are continuous data. So most of the data will be continuous, but questionnaire response will be categorical data. So if it is following normality, what we have to do is, you have to follow parametric test. Parameter means main median, commonly used parameters, main median. If it is following a particular parameter, you can call it as a parametric test. Commonly used parameters are mean, mean and standard deviation. Okay, sorry, we are not made in mean and standard deviation. Meaden will be coming into nonparametric test. So the first test is Z test. Okay, so suppose the sample is less than or sorry, the more than 30, our sample is more than 30 and we are checking or we are comparing a sample with population. Then we use Z test. Z test is very rarely used. Usually we compare two different samples in a study. We hardly compare sample with population. If we compare sample with population, we use Z test. It is a parametric test. If the data is following normality, first you have to think about Z test. If it is more than 30 sample and you are comparing your sample with a population. Okay, if it is not comparing with population, instead you are comparing two groups or before after comparison, you have to follow T test. That is the most commonly seen. So paired is if it is two groups, there is no population here. Just two sample groups, you have to follow paired or independent or unpaired. So this is also known as unpaired test, unpaired test, student data stocks. Okay, so if it is just two groups without any population involvement, anyway, we are comparing the groups, let it be two or three, population scenario comes very hard. So we use two groups, that is T test. So paired means before after comparison, the same group you are applying one particular mouthwash and before you are checking the plaque indices and after you are checking plaque indices or before you give one particular drug for blood pressure before you take the BP and after you take BP, that comes paired T test, that is before after comparison. It is only one group, but they are taken at two point of time before after comparison. So you can count it as two separate groups, even though the group remains same, but before and after observation. But if you have two different groups, group A and group B, you are giving one drug to group A, one drug to group B, and you are checking the difference between them. You can use T test, but should be independent or unpaired or student T test. So that is all about T test, I am not going into very much into the testing, I just want to know on what condition on what scenario each test is being used. So if you have two groups, you can use T test and it should be parametric and it should follow normality. Okay, you have a parameter to measure, that is mean and standard division, that is most common. So ANOVA most commonly used if the study involves more than two groups, so if two groups means you can use T test, more than two groups, it should be ANOVA. So there are many types of, if you have more than two groups, you need to do ANOVA, that is analysis of variance. So if you have a one variable, that is just checking effect of plaque index, effect of, or effect on plaque index, or effect on ginger index, or effect of one particular group, I mean three or four groups, you have to use one way ANOVA, but if you are checking effect of two drinks and drinks may be interactive, then you have to check two way ANOVA, that is two variables are there and you have to make sure that the interaction factor also should be reported. You just cannot do one way ANOVA, if the factors involved in the study are supposed to be interactive, okay, so there can be interaction within the person. So if interaction is there, you have to follow two way ANOVA, two variables and if you have more than two variables, you have to follow MANOVA testing, that is many variables of ANOVA. Then all these will be categorical variables and if you have a continuous variable and a categorical variable, when it involves a continuous and both continuous and categorical variable, you have to use analysis of covariance, that is ANCOVA, you just remember the names, okay, so when you understand the variables and variables better, you get an idea of two way ANOVA and ANOVA and ANCOVA. So what if the data set is not following normality, you have to go for non-parametric test, that is there is no parameter to measure. So there is no corresponding test for the test, but the pre-test, there is one test known as Wilcoxon sign drawing test, that is if you have a before after comparison, you have to use Wilcoxon sign drawing test, it will not check mean and standard deviation, it will be rank the observations and then do the test, that is corresponding to the before after comparison, it is pay-t test, okay. If it is two different groups, not before after comparison, then you have to use a man-witney test, that is corresponding to the independent test, if you have two different groups, you have to follow man-witney test, then this non-parametric test, there is no parameter, usually we take median, which is not a good measure of central tendency, here we take mean and standard deviation, here commonly we take median and Iqor. So if it is a paired observation before after we take Wilcoxon sign drawing test, if it is two different groups, we take man-witney and if it is ANOVA, there is more than two groups, we take Bruce Calvalli's test. So always this parametric test is good, but sometimes we may not be able to do a parametric test since it does not, since your data set is not following normality, you might go for a non-parametric test, there is no much varieties in non-parametric test. So Wilcoxon sign drawing test, man-witney test, these are paired t-test correspondence and man-witney test, parallel to independent t-test and ANOVA from parallel to Bruce Calvalli's test. So here we are checking the difference between two groups, whether the difference you see is due to random chance or a real occurrence, so that is what the statistical testing is done, that's why we are doing statistical testing, we need to find out the difference what we are seeing, whether it is actually happened or it is just by a chance. So what if we are not doing any group comparison, we are just correlating with each other, that is every person in group one will be compared to every person in the group two, so 100 people will be compared to 100 people person by person and we find out correlation, if the data is following normality, we have to follow the Pearson correlation test. Okay, if it is not following normality, we have to follow Spearman rank correlation, here the data is normally distributed, here it is not normally distributed, so you need to know what is correlation before that, so I am just mentioning about the names in this class, not much detail about the testing, I just want you to know on what condition which test is used to, so Pearson correlation is a parametric test which is done for finding out correlation and Spearman rank is non-parametric test, again it is also for correlation, always parametric side will be having better quality, so correlation is not just to find out any significant difference, is how much correlated, if height is going, weight height and weight, if height goes and weight increases, you can say it is positively correlated and if patient is weight is low and BP is high, very rare chance, but still if something is going in opposite direction, if patient is eating too much and reducing his mark or spending too much time on mobile and reducing mark, can say it is negatively correlated, so correlation will be minus 1 to plus 1, if one unit happens, one unit change in the dependent variable, this same change happens in the independent variable or the vice versa, one unit change in independent variable will cause one unit change in dependent variable, you can say that it is perfectly correlated, if it moves the opposite way, one unit change causes negative one unit change in the dependent variable can say it is negatively correlated, so that's all about correlation, so I just wanted to know the test name of correlation, POS and correlation test and Spearman rank correlation test, so that is all about normality, now next we go to the categorical data, we were seeing continuous data, categorical data commonly used as chi square test, this we cannot put into parametric or non-parametric or you can put into both, so it is commonly seen it is known as both the test or neither the test, so is it not? Chi square test always we use four question in this, each question the responses will be in percentage or fractions or proportions, we use chi square test because it is into categories of like a scale options, so in such cases we use chi square test, the parametric or non-parametric test we use commonly based on normality, we use mean standard deviation, median or accurate and here the category is the response percentage for like a scale of five responses, each response 20, 30, 40, 20 in that percentage and many questions, in such cases we use chi square test, so feature exact test chi square test always and 2 by 2, 3 by 3, 4 by 5, 5 by 5 contingency table, in any of the cell if the observation, if the percentage, if the number is less than 5, you have to follow the feature exact test that is low sample size or less than 5 cell count is there, you have to use features exact test which is a type of chi square test and the next one is McNima test, you have a paid data that is case control study before after comparison of proportion or frequency you have to use McNima test, okay so these are the two types of chi square test, one is feature exact test when the sample size is very low or the contingency table any of the cell is having less than 5 number or the McNima test which is a paid data, if the data is paid just like a paid observation in case control study or somewhere or COVID study you can use this McNima test to check the association usually, so the testing is very different each test has to be dealt in very detail but I just want you to know the names of it, okay so feature exact test and McNima test and the last one of any statistical testing is a model prediction that is you have a dependent variable and you have many independent variable you need to know how much change each independent variable is bringing out to the dependent variable, so that is equation prediction equation model prediction, so now we had seen just the difference between two groups or correlation how each group is correlated each other, now you have to know that dependent variable dental anxiety how much it is being predicted by age gender education occupation family income and previous scary previous dental visit, so all these independent variable how it is contributing how much attribute it has for the dependent variable, so prediction equation or model equation will be done by regression analysis, if the data is continuous so the regression model will be linear regression, so finally there will be a r square that will be the outcome of this regression model, so I will be dealing all this in detail about each test but now just remember the name and when this linear and when this logistic regression is being used linear regression is always used in continuous data, okay continuous height weight height weight BP and blood pressure settings, okay logistic regression it should be categories like gender male female we can convert any variable into categorical and continuous data, if suppose if socio-economic status we have we have five categories that is low socio-economic middle upper upper middle, so it becomes categorical data, with the same data we just apply the amount socio-economic data that is we just put the number and we go to kupusami scale and we just use the number 0 to 35 we get a number and we just use the number it becomes continuous data, so either way we can use, so logistic regression or linear regression in logistic regression we will be always seeing odds ratio, here it will be regression coefficient of regression and r square, okay so logistic regression it will be compared to one group other group this much risk or protective effect this will be r square value, so each independent variable will give one r square value that is out of 100 this much percentage is the effect of that particular variable in the dependent variable, okay you just see the names of it I will be explaining in detail what all these things, so continuous data we use linear regression and categorical data logistic regression, so before watching this I I am going to suggest you to watch normal curve and the types of variables, so you will get a better idea after this you will have a faint idea of what test to be done on what condition then once you go through many many articles you will get a very clear idea, okay so that is all about testing, so I was not into detail about testing I was just mentioning the names of it, okay so this is all the test, t-test, ANOVA, correlation, BSM, Wilcoxon, Sintrand test, Manbitt new test, Kruskal-Vales, BM and Rancor relation which is exact test, magnima test, linear regression model and logistic regression model, the first one checking normality, Shapiro-Wilkins test and Smirno-Komogorov test and Anderson-Darlin test okay, so we are coming with a new class, thank you.