Marta García Granero: Every day. I have my collection of macros and if I know that I will need to run something frequently, I turn it into a macro. If I find some special piece of code useful, I add it to the file “macros.sps” and load it every time I start SPSS.

King Douglas: Do you write and/or employ SPSS scripts in your work?

Marta García Granero: No, I'm sorry I haven't had the time to learn yet, but I will. Jan Spousta sent me a really useful one—a reverse auto recode method—that made me think I should definitely learn scripting. I have Ray's book as a starter, and I hope I'll have some time, some day, to start learning. Nevertheless, perhaps I'll devote that time to learning Python, just in case the University finally installs SPSS 14.

King Douglas: What is your comfort level with statistics?

Marta García Granero: Being a Statistics and Experimental Design teacher at the University of Navarra, quite high. I even enjoy reading those boring math sections in scientific papers, and writing programs specifically for certain statistical methods not covered by SPSS.

King Douglas: For instance?

Marta García Granero: For instance, meta-analysis or the Cochran-Armitage trend test.

King Douglas: What are your thoughts about "Instant Statisticians," as made possible by desktop statistical packages?

Marta García Granero: It still astounds me that people who lack any statistical knowledge try to use SPSS with such flippancy. The motto, "Real stats, real easy," is a delusion. Without a solid statistical knowledge, any statistical package can be really misleading…even dangerous.

I remember that, years ago, a student working on her Ph.D. asked me to take a look at her data and results. She thought she had interesting results because there were some significant p-values. She was shocked when I explained to her that the only significant result I could see was due to a failure in the assumption of homogeneity of variances. She thought she had significantly different means, but she had just lack of homogeneity of variances.

I think that sometimes, "when" and "why" a given statistical test should be applied is much more important than "how"…or how easy a certain package makes it. I've seen people saying that they know multiple linear regression because they are able to use SPSS to throw every variable into the predictors box, select "stepwise" and go to have a coffee while the computer builds a model.

King Douglas: Can you recommend a good statistics book for beginners?

Marta García Granero: There is an online e-book I always recommend for beginners. It is called “Statistics at Square One.” I have even written SPSS code (ranging from very simple to quite advanced) to solve every example and exercise presented in the book