Reporting SPSS results

What to keep from the output, and what to leave

The Case Processing Summary, the syntax echo, warnings about empty cells which you have already checked, and the default charts can stay in the output file. Most assumption output also stays there once you have read it. In the chapter, one sentence saying the assumption was checked is enough.

In the thesis you keep the line which answers your objective, along with the descriptive figures the reader needs to follow it. For five common tests, the table below gives the SPSS table you should read and the numbers you should carry into the thesis. Figures in the last column are not real data. We made up a survey of 300 employees just to show how the line looks.

Some scholars take a screenshot of the SPSS table and put it straight into the chapter, which we always ask them to avoid. Type the table again in Word in your university format. Nothing inside an image can be edited, so when your guide wants one small change you are stuck, and to the examiner it looks careless. The SPSS gridlines and font also come along with the picture and they look different from rest of your chapter.

The APA 7 rules that cover most of it

Most Indian universities have not fixed any statistics style for the thesis. If your university also has not given one, APA 7 is a safe choice. The APA’s own numbers and statistics guide is only two pages long. These are the rules scholars need most in a thesis.

• Statistical symbols like M, SD, t, F, p, r and N go in italics. Greek letters such as χ², η² and β are not put in italics.
• Give one space before and after =, < and >. So write p = .03 and not p=.03.
• When a value can’t go above 1, like p, r or α, drop the leading zero (p = .04, r = .42). If the value can go above 1, keep the zero, for example d = 0.60 or M = 0.8.
• Write the exact p-value with two or three decimals. Anything smaller than .001 is written as p < .001. SPSS shows .000, which is not a real value, so never copy that into the thesis.
• Two decimals is normal for most inferential statistics. For means and SDs from integer survey scales the APA guide suggests one decimal, though many Indian theses use two. Choose one and keep it.

As per APA, a figure you have given in the table should not be repeated again in the sentence. If the means are already in the table, the sentence can simply say “permanent staff scored higher” and give the test result only.

Reporting a t-test

In SPSS the Independent Samples Test table has two rows, “Equal variances assumed” and “Equal variances not assumed”. Picking the wrong row is one of the common mistake we are seeing in Chapter 4. First look at Levene’s Test for Equality of Variances on the left side. When its Sig. is .05 or above, take the first row. When it is below .05, take the second row. That second row is Welch’s t-test, and the df will have decimals, for example 171.36.

In recent SPSS versions the effect size for a t-test comes in a separate table called Independent Samples Effect Sizes. From there take the Cohen’s d point estimate. With an older SPSS you won’t find this table anywhere, so you have to work out d by hand, taking the mean difference and dividing it by the pooled SD.

See this example. “Permanent staff (M = 3.8, SD = 0.6) reported higher job satisfaction than contract staff (M = 3.4, SD = 0.7), t(198) = 4.27, p < .001, d = 0.60.” One sentence is enough to carry all of it. Cohen called 0.5 a medium effect and 0.8 a large one, so a d of 0.60 is above medium but below large. Mention that in the text also.

Reporting ANOVA and post hoc tests

Go to Analyze → Compare Means → One-Way ANOVA. Under Options, tick Descriptive, Homogeneity of variance test and Welch. Under Post Hoc, tick Tukey, and also Games-Howell, which you will need if the variances are unequal. In newer SPSS versions the main dialog also has “Estimate effect size for overall tests”. Tick it and SPSS will give you eta squared.

In the thesis, give the F test first. Give the post hoc result after that, and only if F is significant. For example, “Job satisfaction differed across the three branches, F(2, 297) = 6.84, p = .001, η² = .04. Tukey’s test showed that staff in Branch A scored higher than those in Branch C (p = .001), while Branch B did not differ from either.” Notice the two degrees of freedom, between groups first and within groups second.

Now if Levene’s test is significant for your data, report the Welch F with its decimal df and do the pair comparisons with Games-Howell. SPSS prints the Welch F in a small table just below the ANOVA table.

Also explain η² = .04 to your reader in simple words. Roughly 4% of the spread in job satisfaction scores is explained by the branch. Four percent is a small share. Your p-value came significant, yes, but write about the small size also in the same paragraph.

Reporting chi-square and correlation

For chi-square, open Analyze → Descriptive Statistics → Crosstabs. Under Statistics tick Chi-square and Phi and Cramer’s V, and under Cells ask for row percentages. In the thesis you report the Pearson Chi-Square row and put N inside the brackets. “Preference for online banking was associated with age group, χ²(2, N = 300) = 9.48, p = .009, Cramér’s V = .18.” Then refer the reader to the crosstab with percentages, so they can see which group preferred what.

Read the footnote below the Chi-Square Tests table also. Sometimes it says more than 20% of cells have an expected count less than 5. In that case the test is not reliable for that table. You can merge some categories, or for a 2 × 2 table you can report Fisher’s exact test.

For correlation the df is N − 2. “Supervisor support was positively correlated with work–life balance, r(298) = .42, p < .001.” When you have five or more variables, make a correlation matrix. Put the means and SDs in the first columns and show the coefficients only below the diagonal.

Reporting multiple regression

Regression gives three tables and you read each for a different thing. In Model Summary you get R² and adjusted R², that is how much of the outcome your predictors explain when taken together. From the ANOVA table you come to know if the full model predicts any better than having no predictors at all. Coefficients is where you see what each predictor is doing on its own.

In Analyze → Regression → Linear, click Statistics and tick Estimates, Model fit, Descriptives and Collinearity diagnostics. VIF values below about 5 are commonly accepted. Some authors want them below 3.

In the thesis, write about the model first and the predictors after that. The predictors usually go in a table, with B, SE, β, t and p for each one. For the model, a sentence like this works. “The four predictors together explained 28% of the variance in work–life balance, R² = .28, adjusted R² = .27, F(4, 295) = 28.60, p < .001.” Next you pick out the main predictor. “Supervisor support was the strongest predictor, β = .31, t = 5.84, p < .001.”

β is for comparing your predictors with each other. B tells what happens to the outcome when a predictor changes by one unit. Many scholars mix up these two. If your model has paths between several outcome variables, that is structural equation modelling, and our guide on AMOS vs SmartPLS explains it.

Building the tables in Word

In APA style the table number is in bold on one line (Table 4.3), and the title is in italics on the next line. Horizontal lines are used only at the head and the foot of the table, and there are no vertical lines. If there is a note, it comes under the table and begins with the word “Note.” Many Indian universities follow their own caption style instead. There the caption like “Table 4.3: Gender-wise distribution of respondents” sits above the table, and “Source: Primary data” is written under it. Whichever one your template uses, all tables in the thesis should look the same way.

Align the decimals in each column. For the same statistic, use the same number of decimals all through the thesis. When SPSS shows 3.81666, round it to 3.82 yourself and then type, instead of copying the long figure and cutting digits by looking.

The usual layout for a reliability table is given in our guide on Cronbach’s alpha, you can check it there.

The sentences under each table

Under every table in the thesis write two to four sentences, not more. Start by telling in plain words what the test found, and link it to the objective or hypothesis it belongs to. Key numbers can come next, if they are not there in the table already. After that, one line on what the result means is fine. Keep the detailed comparison with earlier researches for the discussion chapter.

In many Chapter 4 drafts you will find “highly significant” or “almost significant” written, and both should be taken out. If the result crosses your α it is significant, otherwise it is not. Write “H2 was supported” or “H2 was not supported”. The line “the null hypothesis is accepted” also has to go, since no test is able to prove the null.

The original output file and the syntax file, keep both with a backup copy on a pen drive or mail. In the viva the examiner may point to some number and ask from where it came, and you should be able to open the file and show him. Our data analysis service can go through the output along with you, and our statistical test chooser helps you confirm the test.