A final parsimonious model (the simplest model that makes an accurate prediction with as few predictor variables as possible) Choose your model based on a balanced assessment of the goodness-of-fit measures (low Akaike information criteria and Bayesian information criteria values), calibration (non-significant P-value for Hosmer–Lemeshow test and visual inspection of calibration plots in the validation sample), and discrimination (high values of AUROC in the validation sample). You then compare the performance of your final model with POSSUM.

There is no automatic process. 1. You should have dependent/outcome variables and predictions. 2. Do the visual binning of predictions. Transform --> Visual binning 3. Aggregate means of outcome variable based on a binned variable. Data --> Aggregate 4. Draw a graph using binned var on X and density on Y. 5. To Draw a line, go to Analyze --> regression --> Curve Estimation (In step#3, I have assumed that your outcome variable is 0 or 1, and hence in this case mean will be the probability of an event happening in that bin).

Hello Dwayne Tucker. First, I suggest you post your question to the SPSSX-L discussion forum, as you are more likely to get an answer there (IMO). A convenient way to post to that forum is via the Nabble archive:

• http://spssx-discussion.1045642.n5.nabble.com/

Click the link near the top of the page for info about how to join the actual mailing list.

Second, if you mean plots like the ones shown here...

• https://blogs.sas.com/content/iml/2018/05/14/calibration-plots-in-sas.html

...I think the following gives a rough approximation:

NEW FILE.

DATASET CLOSE ALL.

* Change path to folder where "sample" datasets are stored.

GET FILE = "C:\SPSSdata\bankloan.sav".

LOGISTIC REGRESSION VARIABLES default

/METHOD=ENTER age ed employ income

/CONTRAST (ed)=Indicator(1)

/SAVE=PRED(PredProb)

/CRITERIA=PIN(.05) POUT(.10) ITERATE(20) CUT(.5).

FORMATS default PredProb (F5.1).

* Chart Builder.

GGRAPH

/GRAPHDATASET NAME="graphdataset" VARIABLES=PredProb default[LEVEL=SCALE] MISSING=LISTWISE

REPORTMISSING=NO

/GRAPHSPEC SOURCE=INLINE

/FITLINE TOTAL=YES.

BEGIN GPL

SOURCE: s=userSource(id("graphdataset"))

DATA: PredProb=col(source(s), name("PredProb"))

DATA: default=col(source(s), name("default"))

GUIDE: axis(dim(1), label("Predicted probability"))

GUIDE: axis(dim(2), label("Previously defaulted"))

GUIDE: text.title(label("Calibration Plot"))

ELEMENT: line(position(smooth.loess.gaussian(PredProb*default)))

DATA: x = iter(0,1,0.1)

TRANS: y = eval(1*x)

ELEMENT: point(position(PredProb*default))

ELEMENT: line(position(x*y),shape(shape.dash))

END GPL.

In SPSS I don't know whether designing adjusted curve is possible or not. However, it is easily possible in Stata with a simple post estimation command. In SPSS, you can draw your function manually as y = exp(bx)/1+exp(bx) https://www.ibm.com/support/pages/plotting-or-overlaying-user-defined-functions-graphs-spss-statistics

Ahmad H. Alomari I have actually seen a few persons suggesting this. I will give it a try and let you know. Many thanks!

Hello Bruce Weaver. Thank you for suggesting the group. It seems like a good community for learning and troubleshooting. I joined immediately posted my question. And yes! I am interested in plots like the ones in your link. I will also try your suggestion. Many thanks!!

Here, I suggest an R code to draw your logistic function manually.

Imagine that your predictor is age ranged from 0-100. One time your coefficient is 2 and in another study is 0.75 as example. You want to compare your coefficients with 1 as a classical logistic function.

Your classical function is exp(x)/1+exp(x). You should correct as exp(x-50)/1+exp(x-50) to align the turning point on the middle of predictor range.

Therefore your code is:

curve(exp(x-50)/(1+exp(x-50)), 0, 100, xlab="Age", ylab="Prabability")

curve(exp(2*x-50)/(1+exp(2*x-50)), 0, 100, col="red", add = TRUE)

curve(exp(0.75*x-50)/(1+exp(0.75*x-50)), 0, 100, col="green", add = TRUE)

A final parsimonious model (the simplest model that makes an accurate prediction with as few predictor variables as possible) Choose your model based on a balanced assessment of the goodness-of-fit measures (low Akaike information criteria and Bayesian information criteria values), calibration (non-significant P-value for Hosmer–Lemeshow test and visual inspection of calibration plots in the validation sample), and discrimination (high values of AUROC in the validation sample). You then compare the performance of your final model with POSSUM.

In order to make calibration plot, you will have to do following steps yourself.

There is no automatic process.

1. You should have dependent/outcome variable and predictions.

2. Do visual binning of predictions. Transform --> Visual binning

3. Aggregate mean of outcome variable based on binned variable. Data --> Aggregate

4. Draw a graph using binned var on X and density on Y.

5. To Draw a line, go to Analyze --> regression --> Curve Estimation

(In step#3, I have assumed that your outcome variable is 0 or 1, and hence in this case mean will be the probability of event happening in that bin).

Full syntax and commands follow.

Creating binned variable:

DATASET ACTIVATE DataSet1.

* Visual Binning.

PRE_2. RECODE PRE_2 (MISSING=COPY) (LO THRU 0.000975873501824482=1) (LO THRU 0.00141701125412073=2) (LO THRU 0.00179052964935964=3) (LO THRU 0.00220329549806634=4) (LO THRU 0.00266698550653581=5) (LO THRU 0.00314263285616621=6) (LO THRU 0.00364267374421248=7) (LO THRU 0.00422690221=8) (LO THRU 0.00488431785475026=9) (LO THRU 0.00571507327812430=10) (LO THRU 0.00727421434905541=11) (LO THRU HI=12) (ELSE=SYSMIS) INTO Binned_pred2. VARIABLE LABELS Binned_pred2 'Predicted probability (Binned)'. FORMATS Binned_pred2 (F5.0). VALUE LABELS Binned_pred2 1 '' 2 '' 3 '' 4 '' 5 '' 6 '' 7 '' 8 '' 9 '' 10 '' 11 '' 12 ''. VARIABLE LEVEL Binned_pred2 (ORDINAL).

EXECUTE.

Calculating Density of event:

AGGREGATE

/OUTFILE= MODE=ADDVARIABLES

/BREAK=Binned_pred2

/outcome_SDH_mean_1=MEAN(outcome_SDH).