# Missing values in SPSS for Chi-Square Tests

I'm a postgraduate student, doing a accent perception study, this is my first time using SPSS!

I'm conducting the above tests on my data, where my participants used Likert scales to analyse the personality traits of 2 guises produced by speakers from Devon. Because of this, some of my responses are 'unknown' or sometimes the participants have put 'don't know' as a response. My question is, do responses such as these count as missing values in my data? Also, is there a specific command in SPSS to ensure that you 'exclude cases pairwise' or is it an automatic thing?

Also, In order to set up my data set on SPSS, I have ordered my data in Excel, before I copy it across to SPSS. For each dependent variable, e.g. 'perceived class' I have 4 speakers producing 2 separate guises and I thought the easiest way to ensure that my scores were above 5 for the chi-square for independence test would be to simply add all the scores per participants for each guise, for example, all 4 of participant 1's 'perceived class' scores for the 2 guises etc...

The Psych textbook I'm using suggests a score of 99 or 999 for missing values, so if I put 99 for a missing value in excel and I am adding those scores together, is it going to affect my scores in Excel?

Just addressing this one issue for now:

The Psych textbook I'm using suggests a score of 99 or 999 for missing values, so if I put 99 for a missing value in excel and I am adding those scores together, is it going to affect my scores in Excel?

You are right to be concerned.

The psych textbook is giving you very bad advice there. In the 1960s, maybe even into the 1970s when there was no suitable alternative it made some sense, and packages (like SAS or SPSS) developed mechanisms to incorporate dealing with numerically-coded missingness.

But such an approach was always risky, and ever since missing values were actually codable in other things, that advice becomes outdated more generally.

In Excel, there's the function NA() which returns the missing value. (try typing =NA() into a cell and see. (But pause after typing the "A" so you can see the description of the function). If you have to do calculations in Excel, you should probably use that.

But if you're going to take it into SPSS you also need to worry about what SPSS can do with missing values; if it has the ability to code missingness non-numerically, I'd suggest you use it. If you have to use a numeric value, you are going to either need to manage numeric missingness in Excel or you need to translate the data as you read it in to SPSS.

You just need to declare the missing value codes in SPSS, which you can do in the Data Editor Variable view (or with MISSING VALUES syntax).

Excel NA values will automatically come across to SPSS as system missing.

Whether user or system missing, such values are generally excluded automatically by the statistical procedures. If you do transformations such as sum, the result is missing if any component is missing when the sum includes any user or system-missing values, but you can use the sum function to sum over all non-missing values.

Do not code missing with a string value in Excel, because that would cause the variable to be read in Statistics as a string.