Binary Logisitic Regression in SPSS with One Continuous and One Dichotomous Predictor Variable

hello this is dr. Grande welcome to my video on conducting a binary logistic regression in SPSS using one continuous predictor variable and one dichotomous variable so taking a look at these fictitious data have loaded in the data editor I have an ID variable and there's a hundred records in the status set I have a continuous predictor named motivation I have a dichotomous predictor variable name referral source and there's just the two levels of course and you have self an agency it's coded as 0 or 1 and then I have a dependent variable also dichotomous and that is completed or not completed again 0 or 1 so we'll assume that this is for a drug and alcohol treatment program the substance use treatment program and there's two outcomes when a client comes into the program they either complete the program or they do not complete the program so that's the dichotomous outcome variable the referral source let's assume for this particular program there can only be two referral sources that would be possible for being referred into the program self and agency so there's no other category here it's a self referral or its agency referred those are the two referral sources so a client can decide to come in on their own or they can be referred by an agency so again dichotomous just two levels there and then motivation would be a score that's observed on a psychometric instrument that measures motivation with a higher score being associated with a higher level of motivation and a lower score being associated with a lower level of motivation before moving into the binary logistic regression because we do have a categorical predictor variable here referral source it's a good idea to run a chi-square and take a look at the relationship between the referral source and the outcome so go to analyze descriptive statistics crosstabs and I'm going to put the presumed predictor variable here which would be referral source in the row list box and the outcome variable in the comm of course it would work either way but typically if we have a predictor variable that goes into the row this box under statistics I'm going to check off chi-square and hit continue under cells I'm going to add expected and the percentages and then click continue no changes for format or style and then click OK and again this is a chi-square not the logistic regression so we have referral source times outcome the cross tabulation and this table we have three tables here at the output but this table the middle table would be of most interest and we can see that when we have a self referral and we look at the outcome there's 17 completed and 44 not completed with an agency referral 35 completed and 4 not completed so it it would appear before even getting into the logistic regression that the individuals referred by an agency have a higher probability of the completed outcome so we keep that in mind as we analyze the results from the binary logistic regression so to conduct the binary logistic regression we'll go to analyze regression and binary logistic notice that I'm running the statistic from the statistics viewer it can be conducted from the data editor as well so here is the dialog for logistic regression the dependent variable would be the outcome and motivation is considered a covariate here and in the logistic regression dialogue and the referral source is considered a covariant so these are the predictor variables now one is continuous motivation but referral source is nominal in this case dichotomous so I'm going to click categorical and move referral source over to the categorical covariance list box notice that by default the reference category is set to last I'll click continue under save I'm going to save the probabilities and the group membership these two will be new variables on the data editor this will not appear in the output the probabilities and the group membership these are both variables click continue under options and I'm going to add the classification plots the HL goodness of fit and the confidence interval for the xB beta and click continue and there'll be no changes under style just click OK and we can see here the results from the logistic regression we have no missing cases the dependent variable encoding completed is zero not completed as one for referral source self and agency we have 61 self-referrals 39 agency referrals then moving down the output tables I want to get down to model summary we can see here our square 0.59 5 so roughly 60% of the variance in the dependent variable is explained by the predictor variables we also have a non-student statistically significant result 0.9 63 on the HL test and that's what we would want here then moving down the output tables to the classification table here we have the observe to outcome and the predicted outcome and we can see that most of the time this model predicted the correct outcome for completed observation and completed outcome we have 43 not completed for predicted just nine and four not completed and the completed predicted category we have eight and an observation of completed that was in fact predicted as not completed 40 so the model appears to be working fairly well just looking at the classification table and then we want to interpret the variables in the equation and a particular interest first would be the p-values and we can see here that we have a statistically significant finding for motivation and for referral source level of referral source would be the self referral and before we interpret the exp beta I want to take a look down here we could see the predicted probability is of membership for not completed so we keep that in mind as we interpret variables in the equation so we look at motivation what this tells us that as motivation increases you can see we have a point nine one four value here as motivation increases the odds of outcome being not completed decrease and more specifically if we take one and subtract this value it gives us point six or eight point six percent so as motivation increases by one unit the odds of a not completed outcome decreased by eight point six percent when taking a look the referral source and again this is for self referred the seven point two seven one tells us that when we have a referral source of self the odds that we're going to have a not completed outcome are increased by seven point two seven one times it's seven point two seven one times higher for the self referral as compared to the agency referral and then the last area one take a look at is the data editor I'm going to take a look at the two variables that were created one is the predicted probability and the other is the predicted group and again these are on the data editor and not on the output and this first variable predicted probability gives us the probability that this particular case this particular record will have an outcome of not completed so for this first record with a motivation of 42 and a self referral there was a 70% probability that not completed be the outcome you can see in this case the model was incorrect the actual outcome was completed and the predicted group was not completed in the next record however it was correct with motivation at level 58 and the referral as self referred the actual outcome was completed the probability that the outcome would be not completed was only about 36% so the predicted group was completed I hope you found this video on conducting a binary logistic regression and SPSS to be useful as always if you have any questions or concerns feel free to contact me I'll be happy to assist you

Channel: Dr. Todd Grande

Views: 36,962

Rating: 4.9203186 out of 5

Keywords: SPSS, binary logistic regression, multinomial logistic regression, logistic regression, regression, dependent variable, independent variable, predictor variable, continuous, continuous variable, scale, dichotomous, dichotomous variable, exp beta, exponentiated, beta, confidence interval, outcome variable, variable, model fitting, odds, odds ratio, probability, group membership, predictor, hosmer-lemeshow, covariate, chi-square, p value, predicted group, counseling, Grande

Id: A0xEwQyN2sE

Length: 10min 19sec (619 seconds)

Published: Mon May 23 2016