ANOVA is an extended t-test for 3 groups or more. Mental well-being is better than efficacy of therapy because you have one group who never had therapy and then you have face-to-face therapy and online therapy.

Compute a new grouping variable (e.g. GROUP) and use this as grouping variable for your independent t-test .

COMPUTE GROUP = 1. *so you set it on 1 for all participants

IF (Q1 = 2 AND Q2 =2) GROUP = 2. *here you set all participants who answered both questions with "no" on value 2

EXECUTE.

Now the grouping variable GROUP has a 1 for all who answered 1/1, 1/2 or 2/1 and a 2 for all who answered 2/2 in Q1 and Q2. Is that what you wanted?

P.S.: generally I would suggest to use 0/1 coding for dichotomous variables, in case if you ever want to conduct regression analyses.

Try using MS Excel to produce the new variable and import the data back into SPSS for analysis.

Also, I think your response variable "perceived efficacy of online therapy" is more of a score and therefore cannot be considered as a continuous quantitative variable. And in this case, the student t-test is no longer applicable. I therefore suggest that you check the nature of your response variable to be able to use a more appropriate method.

I do not agree with Mugumaarhahama Yannick there is no need to switch to another program, if not necessary. Try my syntax code and tell us if it worked please.

Mugumaarhahama Yannick why do you think that a rating of perceived efficacy can not be continous? Why cant a score be continous, does this contradict each other (I dont think so...)?? You are right that everyone should check assumptions and the measurement scale of one's variables, but I do not see a general problem in Marc's description.

ANOVA is an extended t-test for 3 groups or more. Mental well-being is better than efficacy of therapy because you have one group who never had therapy and then you have face-to-face therapy and online therapy.

Follow answer.

Following answers

Hi guys! Thank you all so much for your kind replies! Rainer Duesing, what you’re suggesting sounds exactly like what I want to do, but I’m not sure how to navigate SPSS in order to apply it. I tried several things, amongst which the following. So, the two questions of interest (face-to-face therapy and online therapy) are dichotomous variables with YES (=1) or NO (=2) answers. What I did first was create 3 new variables defined as follows: (1) face-to-face_therapy=1 AND online_therapy=2; (2) face-to-face_therapy=2 AND online_therapy=1; (3) face-to-face_therapy=1 AND online_therapy=1. Then, I added those 3 variables into a new, bigger variable with the syntax (1) AND (2) AND (3). For some reason, that doesn’t seem to work; when trying to apply the t-test, I can't do the grouping anymore because I end up with two separate variables (psychotherapy=YES and psychotherapy=NO), and SPSS asks to group by one variable only with two groups within it (rather than two separate variables of one group each, if that makes sense). What I would need is a way to integrate all of the YES/NO answers into a new dichotomous variable (one variable with YES/NO groups inside of it), but I'm not sure how to do that.

s. Béatrice Marianne Ewalds-Kvist, I’ve thought of using ANOVA, but that’s not the ideal solution because I’m not interested in the partial results of each therapeutic modality; rather, what I want is to collapse all forms of therapeutic contact (face-to-face YES, online NO; face-to-face NO, online YES; face-to-face YES, online NO) into one overarching category (psychotherapy YES), and then contrast that with the group that’s never had any kind of psychotherapy. We do have data on mental well-being as well, though the main goal in this bit of research is to figure out social perception of online therapy; my hypothesis is that people who’ve had any experience of psychotherapy at all will be more trusting of online therapy, and have more positive expectations of it, than the group that has never undergone any kind of therapy. That’s why the dependent variable is “perceived efficacy”. Mugumaarhahama Yannick, the dependent variable is on a quantitative scale ranging 0-40; I guess your point is that only whole numbers are accepted as answers, so maybe that means I wouldn’t be able to apply a t-test or ANOVA?

Dear Marc Josep Garcia Hervás

My concern was only with your response variable. Does it take real values (belonging to IR) or integer values (belonging to IN). If it takes real values, does it follow a normal distribution?

The answer to these questions should guide you in choosing the statistical method to use.

What do you mean with "navigate SPSS"?!?! You open a new syntax file and copy paste this code (as already described above):

COMPUTE GROUP = 1.

IF (Q1 = 2 AND Q2 =2) GROUP = 2.

EXECUTE.

Done! Nothing more to do, you should have a new variable "GROUP" in your variable view.

If you did it as you described it above it is cler why it did not work. The word "AND" is a logical operator, which defines that two conditons have to be met to do something, it does not add up anthing, so an AND is very different to a SUM.

Rainer Duesing, thanks for getting back to me! It's my first time using SPSS and I don't know many of its functionalities. I've copied the syntax you suggested and run the script; it does create a new variable named GROUP, but it doesn't have any values in it that I can see in "variable view". Even if I go ahead and set values of 1 = YES and 2 = NO, when I run the t-test (using efficacy as the contrasting variable and GROUP(1 2) as the grouping variable) I get an error that says that "T cannot be calculated because at least one of the groups is empty", that group being "NO" (value = 2). The data file has hundreds of variables -- how can I select the right ones for GROUP 1? I get the "IF (Q1 = 2 AND Q2 =2) GROUP = 2" command, but not the "COMPUTE GROUP = 1".

Mh, I dont know what you did, but the code works perfectly fine on my machine(s), see screenshot.

The Idea behind the COMPUTE command is, that you create a new variable with only 1s in it, for every participant. The is parsimonously, since 3 of your 4 conditions will end up with a 1 anyways, so I assign 1s to everyone, instead of writing 3 IF commands. Only for the fourth conditon Q1 and Q2 equal 2 I will use an IF command and assign 2s to those people. 3 lines of code instead of 5.

Again, have a look at the screenshot, you can see that the GROUP variable is correct for every combination of conditions.

Firstly, recode all the variables (Q1 and Q2) to 1 (yes) and 0 (no). Create a new variable (say Total) using Compute command in SPSS as follows:

COMPUTE Total = Q1+Q2.

The Total variable will have three values:

0: No therapy

1: Either ONLINE or FACE-TO-FACE therapy

2: Both therapy.

I hope this will be helpful.

Hey guys! Sorry for the late reply -- I was wrapping up the paper. I wanted to thank you all, and especially Rainer Duesing, for your patience, kind help, and invaluable input. Rainer Duesing, your script did work in the end; this is painfully embarrassing to admit, but apparently the reason the syntax was not working at first is that I had to select it (as in, have it highlighted in blue) before executing the command (which I didn’t know, this being the first time I used SPSS). Again, thank you so very much!

You are welcome! What matters is that it worked in end! All the best for your publication.