RESEARCH NOTES
HOW TO ANALYSE YOUR RESEARCH DATA? ILLUSTRATIONS WITH HANDS-ON EXERCISES USING SPSS.
The SPSS commands for obtaining the statistics ‘HbA1c’ breakdown by ‘Race’ in Table 1 are as follows:
From the menus choose:
- Analyze
- Reports
- Case Summaries
- Select ‘hba1c’ into the ‘Variable’
- Select ‘race’ into the ‘Grouping Variable’
- Click on the icon ‘Statistics’ and select ‘Number of Cases’, ‘Mean’, and ‘Standard Deviation’
- Uncheck these two boxes ‘display cases’ and ‘Limit cases to first 100’
- Click ‘OK’
Type of data
The first thing to take note is the type of data (or variables) you have collected.
- Categorical data. There are basically two kinds of data in this groups:
- Nominal data (named categories), e.g. gender (male/female), ethnicity (Malay, Chinese, Indian), outcome (dead/alive), etc. The nominal data are summarised by percentages.
- Ordinal data (ordered categories), e.g. tumour staging (Stage 1, 2, 3, 4), disease severity (mild, moderate, severe), Likert scale (5-point scale, 1-5), etc. The ordinal data are summarised by median value.
- Continuous data. Continuous data is sometime referred to as interval data. These data take the form of a range of number, and may or may not have decimals, e.g. age, HbA1c, weight, height, haemoglobin level, etc. The continuous data are summarised by mean and standard deviation (SD).
Another way of looking at the data is defining the dependent and independent variables:
- Dependent variable is the variable of interest
- Independent variable is the grouping variable
Summarised data
We have seen already earlier different type of data are summarised differently. This summary of data, plus a graphical display of the data (e.g. in graph and scatter plot) is a very useful way of having a sense of your data before you embark on formal statistical analysis (the so-called “eye-balling the data”).
Hypothesis testing, sample and population
One of the reasons for conducting the above study is that you have observed that diabetic patients of certain ethnic group appeared to have poorer diabetic control. Rather than stating that “Malay diabetic patients have poor diabetic control”, we should state that “in the population of all diabetic patients, there is no difference in glycaemic control by ethnicity” (this is the so-called Null Hypothesis). By drawing a representative sample of diabetic patients from the population, you then seek to disprove the Null Hypothesis. This process of drawing conclusion on a population from a sample is called inferential statistics.INFERENTIAL STATISTICS: PARAMETRIC TESTS
t-test
If you want to find out if HbA1c differ by gender, a statistical output can appear as follow: means HbA1c for males and females are 8.7% (SD=1.9) and 8.9% (SD=2.3) respectively, t= - 0.711, df=158, p=0.478. As HbA1c is a continuous variable (and presumably normally distributed), we use t-test for two groups comparison of means (males vs females). Since the p value is more than 0.05 (the conventional cut-off for statistical significance), we can interpret the result as no statistical significant difference or “no real difference in HbA1c in male and female diabetic patients”.
The SPSS commands for obtaining the above t-test statistics are as follows:
From the menus choose:
- Analyze
- Compare Means
- Independent-Samples T Test
- Select ‘hba1c’ into the ‘Test Variable’
- Select ‘sex’ into the ‘Grouping Variable’
- Click on the icon ‘Define Groups’
- Type in ‘1’ in ‘Group 1’ and ‘2’ in ‘Group 2’
- Click ‘OK’
ANOVA
For three or more groups comparison of means (Malays, Chinese and Indians), we use ANOVA (F test): means for HbA1c in Malays, Chinese and Indians are 9.6% (SD=2.5), 8.4% (SD=2.0) and 8.7% (SD=1.9) respectively, F=4.524, p=0.012. In this case, there is statistical significant difference (p is less than 0.05) in the HbA1c among these three ethnic groups.
The SPSS commands for obtaining the above F-test statistics are as follows:
From the menus choose:
- Analyze
- Compare Means
- One-Way ANOVA
- Select ‘hba1c’ into the ‘Dependent List’
- Select ‘race’ into the ‘Factor’
- Click ‘OK’
Note: One-Way ANOVA is chosen because there is ‘one factor’ (i.e. ethnicity) in the example. Although there are three ethnic groups in the study, it is considered one factor because all these groups can be grouped into one factor i.e. ‘ethnicity’ for analysis.