Multiple regression with bootstrapping in SPSS

in this video I'm going to show you how to use bootstrapping as part of the multi regression analysis the data sent here I'm using is the same as I used in the video for the one-way ANOVA you can see here we have five different variables the first one is student score on the standardized last test the second one is their self-efficacy those variables are continuous and the difference is for the math self-efficacy is a standardized variable the red play with a mean of zero a standard deviation one and the next two variables are the hours per week students dead maths homework in school and math homework out of school last one is student answer to this question mathematics is important their answers are on the 4 point Likert scale from strongly agree to strongly disagree okay so go back to the data view if we are going to run the mall progression here's what we do can analyze regression linear I'm going to use the standard score as a dependent variable self-efficacy and the hours work in and out of school math work as the independent variable so when I run a mock regression I always want to have the descriptive information as well as the Kohner narrative next gnostics at this point I'm not going to do anything with describing because again I want to compare the same analysis with and without boost reading so you can see the difference you click on OK and the output is here you can see that for all four variables the one dependent variable on three independent variables I have number of observations 100 and you're given the mean and standard deviation for each of the variables and next table is the Pearson correlation between all pairs of variables you use in the study whether they're statistical significance okay and you can see that this model explained a significant significant de Monterrey's in the dependent variable and the variables that contribute significant the math war according to the test of significance is the student self-efficacy and the amount of time they spent on homework math homework out of school okay and here you're given the coordinate architects classic but you can tell from here we do not have any problem with um multicollinearity all right now we're going to run same analysis in the same linear regression place the only difference is we're going to require additional information by using bootstrapping ok I'm going to take the deep default 1000 a number of samples and ask for bias corrected accelerate estimate for the confidence interval and leave the rest as the way they are okay and we are just only to do is click on OK and you have this part of the output now you can see that in addition to the original the mean and standard deviation sample size information we're getting the information from bootstrapping estimates so for example for the first variable the dependent variable the mean in 35 and based on the bootstrapping estimate the confidence interval 95% confident confidence interval for this mean is fifty three point two two fifty six point five so this 95% call from the interval is offered not only for me but also for the standard deviation and the same is available for all variables in this analysis okay move forward we can see that in the correlation table we're still provided with the bivariate Pearson correlation between all pairs of the variables in this analysis along with the test of significance and the valid number of observations there are more here we start from this point the split striving for Pearson correlation the first part we are offered the bias and based on the bias we have the estimate for that error and this confidence interval for each of the Pearson correlation okay so um first part you're given the lower bound is the heart rate because if you want to look for the opera ball you have to move far down here for the upper bound so for example the Pearson correlation between the standard math score and the self-efficacy is point two nine eight point two eight nine that's four lower bound and upper bound is point six three zero so on so forth alright now after the Pearson correlation we're given the information about the regression model that's the same as a regular output this model explained significant amount variance in the dependent variable and according to the test two of the independent variables are statistically significant they maths efficacy and hours per week spent maths or more out of school okay okay now we have another table that's called put foot strap for coefficient because the regression coefficient are the parameters we needed for a estimate of a regression model in the books tracking of estimate for each regression coefficient there is a bias estimate provided and also the 95 confidence interval for that particular parameter altogether you can see that when we request boost tracking in the multiple regression all the important parameters or sample statistic were interesting including the mean of the variables the correlation between variables and the regression coefficient we are gaming the most rugged estimate which is based on one solid samples repeatedly taken from the original sample so the boost reading estimate provided here you know the confidence interval is supposed to be more reliable and more accurate than the information we had from one single sample especially one this when a model sometimes violate the underlying assumption like normality homogeneity of variance so this is how we take Vantage of bootstrapping in regular analysis

Channel: Soft6Stone

Views: 18,758

Rating: 4.1020408 out of 5

Keywords:

Id: giffY-hScuY

Length: 8min 29sec (509 seconds)

Published: Mon Apr 03 2017