Finally, for Socio Economic Class ( sec) we will use the least affluent class as the reference (‘Never worked/long term unemployed - 8’). Previously we have used male as the reference so we will stick with this (once again, change the selection to ‘First’ and click ‘Change’). For the Gender variable we only have two categories and could use either male (‘0’) or female (‘1’) as the reference. Change the selection to ‘First’ and click ‘Change’. For our Ethnic variable the first category is ‘0’ White-British (the category with the highest number of participants) so, as before, we will use this as the reference category. 6 ).Īll we need to do then is tell SPSS whether the first or last category should be used as the reference and then click ‘Change’ to finalise the setting. For our purposes we can stick with the default of ‘indicator’, which essentially creates dummy variables for each category to compare against a specified reference category – a process which you are probably getting familiar with now (if not, head to. This allows you to alter how categories within variables are compared in a number of ways (that you may or may not be pleased to hear are beyond the scope of this module). The first thing to note is the little drop down menu which is set to ‘Indicator’ as a default. To do this we must click on each in turn and use the controls on the bottom right of the menu which are marked ‘Change Contrast’. The next step is to tell SPSS which category is the reference (or baseline) category for each variable. You need to move all of the explanatory variables that are categorical from the left hand list (Covariates) to the right hand window… in this case we need to move all of them! To do this we need to click the button marked ‘Categorical’ (a rare moment of simplicity from our dear friend SPSS) to open a submenu. Now they are there we now need to define them as categorical variables. To start with, move ethnic, SEC and Gender into the covariates box. If the explanatory variable is continuous it can be dropped in to this box as normal and SPSS can be trusted to add it to the model, However, the process is slightly more demanding for categorical variables such as the three we wish to add because we need to tell SPSS to set up dummy variables based on a specific baseline category (we do not need to create the dummies ourselves this time… hooray!). This variable is labelled fiveemand should be moved in to the Dependent box.Īny explanatory variables need to be placed in what is named the covariates box. Our outcome measure is whether or not the student achieves five or more A*-Cs (including Maths and English) and is coded as ‘0’ for no and ‘1’ for yes. First of all we should tell SPSS which variables we want to examine. The logistic regression pop-up box will appear and allow you to input the variables as you see fit and also to activate certain optional features. Take the following route through SPSS: Analyse> Regression > Binary Logistic You can also follow the process using our video demonstration. We will create a logistic regression model with three explanatory variables (ethnic, SEC and gender) and one outcome ( fiveem) – this should help us get used to things! You can open up the LSYPE 15,000 Dataset to work through this example with us. Let’s get started by setting up the logistic regression analysis. To do this we will need to run a logistic regression which will attempt to predict the outcome fiveem based on a student’s ethnic group, SEC and gender. To gauge the extent and significance of any interactions between the explanatory variables in their effects on the outcome.Specifically we are interested here in what the OR for ethnicity looks like after we have controlled for differences in exam achievement associated with SEC and gender. ethnicity) on the outcome variable when controlling for other variables also associated with the outcome (e.g. To gauge the effect of one explanatory variable (e.g.Are the effects in the sample sufficiently large relative to their standard errors that they are likely to be true in the population?
Remember that this data represents only a sample (although a very large sample) from the population of all students in England (approximately 600,000 students in any one year group). To evaluate the statistical significance of the above associations.Why would we want to get involved in logistic regression modelling? There are three rather good reasons: So we can see the associations between ethnic group, social class (SEC), gender and achievement quite clearly without the need for any fancy statistical analysis.