Hi Cristian. Do you really need a new variable? Or would the results from the CROSSTABS command be sufficient?

• http://psychology.illinoisstate.edu/jccutti/psych340/fall02/oldlecturefiles/crosstabs.html

If you really do need a new variable (e.g., for use as an outcome variable in a regression model), try this (changing variable names as needed):

COMPUTE Both = (orange EQ 1) AND (fplugs EQ 1).

FORMATS Both (F1).

VARIABLE LABELS Both "Orange color & follicular plugs".

VALUE LABELS Both 0 "No" 1 "Yes".

FREQUENCIES Both.

This assumes 1 = Yes for both variables. If both variables are coded such that 1=Yes and 0 = No, you can simplify the COMPUTE command, like this:

COMPUTE Both = orange AND fplugs.

HTH.

Dear Bruce,

Thank you very much for your quick response. Very useful! Your formula worked perfectly!!

Actually, that's a good question. I analyzed both variables independently:

Presence of orange color (yes/no) associated with cases/controls

Presence of follicular plus (yes/no) associated with cases/controls.

I came out with the idea to see how often both are together but don't know if that adds something...

What do you think?

Suppose you have

X = OrangeColor = 0 if orange color is not present

X = OrangeColor = 1 if orange color is present

Y = FollicularPlugs = 0 if follicular plugs are not present

Y = FollicularPlugs = 1 if follicular plugs are present

Define a new variable

Z = 2 * X + Y

then,

Z = 0 if X = 0 and Y = 0 (no orange color and no follicular plugs)

Z = 1 if X = 0 and Y = 1 (no orange color and follicular plugs)

Z = 2 if X = 1 and Y = 0 (orange color and no follicular plugs)

Z = 3 if X = 1 and Y = 1 (orange color and follicular plugs)

The frequency of Z = 3 says you how frequent BOTH variables are present simultaneously.

if you are only interested in how frequent BOTH variables are present simultaneously, you may define

U = X * Y

U = 1 if BOTH variables are present simultaneously

U = 0 elsewhere

best regards

Dear Jorge,

Thank you so much for your answer.

Kudos to Researchgate for creating this incredible platform and to both of you for helping me!!

Best regards,

Hello Cristian. You mentioned cases & controls. It seems to me that you want a logistic regression model with Case (1=Case, 0=Control) as the outcome variable, and with Orange, Fplugs and their interaction as explanatory variables. Including the interaction term allows you to directly compare the effect of one variable at both levels of the other. You could use the LOGISTIC REGRESSION command to estimate this model, but it is arguably more convenient to use GENLIN with a logit link function and binomial error distribution. GENLIN has an EMMEANS sub-command, which is useful for decomposing the interaction term. I don't have SPSS on this machine, but I believe the syntax would look something like this:

GENLIN case (REFERENCE=FIRST) BY orange fplugs

/MODEL orange fplugs orange*fplugs

INTERCEPT=YES

DISTRIBUTION=BINOMIAL LINK=LOGIT

/EMMEANS TABLES=orange*fplugs SCALE=TRANSFORMED COMPARE=orange

/EMMEANS TABLES=orange*fplugs SCALE=TRANSFORMED COMPARE=fplugsANSFORMED COMPARE=WHITE

/MISSING CLASSMISSING=EXCLUDE

/PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION (EXPONENTIATED).

The tables generated via EMMEANS will show the fitted log-odds of being a case for each of the 4 cells you get by crossing your two explanatory variables (orange x fplugs). The COMPARE options will give you the differences in log-odds for the simple main effects. You can get odds ratios by exponentiating those differences in log-odds. An easy way to do that is to copy the table (as Excel), paste into Excel, then use the EXP() function in Excel.

HTH.

Hello Cristian

1. The above explanations by colleagues are useful.

2. It may be very useful to follow the attached documents.

Regards,

Zuhair

Go to compute variables and combine them through by taking their mean or sum!