Discriminant analysis using SPSS: By G N Satish Kumar

you discriminant analysis in this video we will be discussing about performing discriminant analysis discriminant analysis and it takes our same task as multiple linear regression the major difference is indiscriminate analysis the dependent variable will be categorical and independent variables can be scale data or ordinal data I am taking example in which a half confident of the employee as dependent variable and age experienced expert and satisfaction as independent variables we have asked her employee are you confident to do the given job the respondent answered as one or two one says not confident to do the given job too confident to do the given job right we have recorded the age of the employee experience of the employee in yes and expertise of the employee and satisfaction as ordinal expert it is asked I am expert in my job the respondent can give the answer and Likert scale from one to five one is strongly disagree five is strongly I agree satisfaction I am satisfied with my job strongly disagree to strongly agree now basing on these four variables age experienced expert in job and satisfaction we want to see as a conference can discriminate these four variables I am starting the analysis to do the analysis we go for the menu and I'll analyze in this classify in classify we go for discriminant right as I already said discriminant analysis the categorical variable will be dependent so I am taking confident as a grouping variable and defining the one minimum maximum two and I taking the other four variables as independent variables now I am going to do small settings in statistics menu and classify first I do for statistics in statistics check button may means you need variate ANOVA box plot select this check buttons and in classify leave one out classification and after doing this to setting statistics and classify say okay now this is output screen total number of respondents are 384 I don't have any missing values in this and when I comfort discriminant I want to see the difference between not confident too confident in terms of four parameters for variables age experience expert and satisfaction I want to see how far they are very ating mine all hypothesis will be always saying that there is no variation between not confident group to confront group let me check in terms of age the mean is 26 in this case the mean is 31 so they are different when you come for experience the mean in not confident is 2.9 for whereas here it is 7.17 means are different when you come for expert it is 4.05 but here it is 4 point 4 0 means when an employee saying I am expert in doing a job this is not much discriminating this variable is not discriminating satisfaction not confident 3.8 2 and here it comes four point three seven so out of four variables when you see the means age is discriminating experience is discriminating expert is not much discriminating satisfaction is discriminating let me see boxplot test you can see these test results here in box plot test the null hypothesis says equal population covariance matrices it mean to say there is no difference between car friend and not car front group we need to check significant value if this value is less than 0.05 we reject the null hypothesis saying that there is a difference between not confident to confront group okay so here we are rejecting the null hypothesis of equal population covariance matrices let me go for canonical discriminant functions when you are saying canonical discrimination we need to check this parameter canonical correlation the canonical correlation is 0.82 when we take square of this value it shows the variance between the groups okay canonical correlation is zero point eight two four let me do the square of a zero point eight two four it is 0.67 that is around sixty seven percent variance is been shown between not conference to confront group it is good it is more than fifty percent variance we are able to see so when we see canonical correlation which is zero point eight two four take the square of that that shows the percentage here in this case it is 0.67 so it is 67.8% of variance is being shown to mean not car friend to confront group when you come for Wilks lambda we can see the significance is zero point zero zero means it is significant if it is less than 0.05 we take it is significant we need to see Wilks lambda values zero point three to one Wilks lambda will range between zero to one if this Wilks lambda value is close to zero indicates that group means are different if the value is close to 1 there is no different now we can conclude it is there is a difference between the group means of non not car friend group to confront group so I have shown you three types of tests one is between the groups I have clearly shown second one is boxplot M test varied to see a significance value third one is Wilks lambda so with this three tests we come to a conclusion that discriminant analysis is discriminating between not car friend to confront group there is a variance of 67% okay let me see like a fact all I know of multiple regression let me see these values H is zero point five six six experience is zero point five six five and expertise in job is 0.4 this is less impacting factor whereas age experience is better impacting the discrimination followed by satisfaction expertise is not much impacting if you want to say the importance of variables first importance is through H it is 0.56 six followed by experience 0.565 third one is satisfaction expertise is very less it is 0.046 this is what we can come to a conclusion here if discriminant analysis is discriminating then age experience satisfaction are three variables which are playing major role and expert is not playing much important role now let me say a very important thing this is called classification results and it is famously called as confusion table in confusion table we need to see at this point which is ninety seven point two one percent this is called hit ratio hit ratio explains how much discrimination is happening for example you can see here not confident people are predicted as not confidence 233 means if you see percentage-wise original not confident is predicted as not confidence 97.5% very high person showing people originally or who are confident are predicted as current is ninety six point six percentage totally the discriminatees there is a hit ratio of 97.1% it mean to conclude the discrimination is being discriminating the total response into not confident and comfort employees basing on four important parameters or variables age experience expert and job satisfaction at a variance of 67% and the hit ratio is 97.1%

Channel: My Easy Statistics

Views: 33,093

Rating: 4.6945453 out of 5

Keywords: Multivariate Data Analysis, Multivariate Analysis, SPSS, Discriminant Analysis, Cluster Analysis, Two Step Cluster Analysis, Hierarchical Cluster Analysis, two step cluster analysis spss, two step cluster analysis spss output interpretation, two step cluster analysis in r, discriminant analysis in spss, big m method, discriminant function, discriminant analysis using spss *****, multivariate data analysis lecture, multivariate data analysis in r, multivariate data analysis spss

Id: LMNsEaC0rAY

Length: 13min 42sec (822 seconds)

Published: Thu Aug 04 2016