VIEWS: 12 PAGES: 7 POSTED ON: 4/1/2011
UCL INFORMATION SERVICES DIVISION INFORMATION SYSTEMS Using SPSS: Tasks If you are completing these tasks as part of a training course the files can be found in the following folder: spss Please save all your work in this folder during the training session. Files can also be downloaded from www.ucl.ac.uk/isd/common/resources where you will also find copies of these tasks and the accompanying manual for download. Exercise 1: Defining variables and entering data You are responsible for collecting data from a clinical trial of Drug X. Drug X is postulated to affect blood levels of a certain hormone (hormone H), so levels of the hormone will be measured before and after treatment with X. In addition to the hormone data, five other pieces of information will be collected from each participant in the trial. 1. Switch to the Variable View and define the seven variables listed in the table below. Use numeric variables in SPSS for categorical data. Variable name Data type Surname Gender Categorical (categories are Female and Male) Age Continuous Income Continuous Smoker Categorical (categories are Smoker and Non-smoker) Hbefore Continuous Hafter Continuous 2. The table below shows the data which you have collected from five patients who took part in a pilot study for the clinical trial. Switch to the Data Editor and enter the data on these five patients. Remember to use numeric codes where necessary. Surname Gender Age Income Smoker Hbefore Hafter ROBBINS Female 32 46000 Non- 94.58 88.79 smoker MCGREGO Male 33 58000 Non- 106.12 78.25 R smoker KUMAR Male 38 47000 Smoker 88.11 102.45 ALLINSON- Female 51 55000 Non- 83.62 63.82 HENRY smoker OLDER Male 44 28000 Non- 72.31 77.50 smoker Save the data as pilotgroup.sav. Using SPSS: Tasks 1 September 2008 Exercise 2: Variable and value labels Open the data file medicaltrialX.sav in SPSS. 3. Create the following labels for the variables: income Household income smoker Smoker or non-smoker hbefore Blood levels of H before treatment hafter Blood levels of H after treatment 4. Create the following value labels for the gender variable: 1 = Female 2 = Male 5. Save the file with the new definitions. Exercise 3: Missing data Some of the data in medicaltrialX.sav contains missing values. 1. Inspect the data in your data sheet and spot any missing values. Take note of which variables have missing values. 2. For each of the variables identified above, decide on an appropriate coding for missing values. For example, for a numeric variable which cannot be negative, -1 might be used. For text data, an X might be used. 3. In Variable View, specify missing values for each variable identified in paragraph 1. above. 4. After specifying the missing values, return to the Data Editor and change any blank cells to the appropriate missing value code. 5. Run the Frequencies command which shows that the missing values have been coded. (Hint: Use the menu option Analyse | Descriptive statistics | Frequencies, move all the variables into the right-hand box, then click OK.) 6. Close the output window (do not save the output). Save the changed data file, giving it the new name fixed.sav . Exercise 4: Importing an Excel file 1. Start a new, blank data sheet. 2. Import the file results.xls into SPSS, using the column headings as variable names. 3. Switch to Variable View. Add a new variable called SEN, and give it the label “Special Educational Needs”. 4. The SEX variable (as in an earlier exercise) uses values of 1 and 2 to indicate “Female” and “Male” respectively. Enter the value labels for the SEX variable. 5. Save the file to results2.sav. Using SPSS: Tasks 2 UCL Information Systems Exercise 5: Sort Cases and Select Cases Load the data file medicaltrialX.sav. 1. Sort the data in order of age (oldest first). 2. Sort the file in order of smoker within gender (i.e. gender is the primary ordering). 3. Select all the males in the group. 4. Select “all cases” again. 5. Now select the patients whose hormone levels were greater after the treatment than they were before. There is no need to save the data file at this point. If you have extra time, see if you can select all the smokers. Exercise 6: Recoding variables Helpful hint: If you still have a subset of the cases selected (from the previous exercise), make sure you select all cases before you proceed. We need to recode some of the data in medicaltrialX.sav (or fixed.sav). The information in the smoker column is coded as text (Y and N). It would be better to code it as numeric data (e.g. 1 and 0), and to use labels to indicate the meaning of these numbers. 1. Use Recode to convert the smoker information into a new variable called smoker1, so that 'Y' becomes 1 and 'N' becomes 0. (Define the label for this new variable as "Smoker or non-smoker") 2. Create value labels for the smoker1 variable so that 1 is displayed as 'Smoker', and 0 is displayed as 'Non-smoker'. 3. Check that the recode has worked properly. If it has, then delete the old variable smoker from your data sheet. 4. Using Recode, create a new variable incband (with label "Income band") to categorise the household income: up to $25,000 as band 1, between $25,000 and $40,000 as band 2, and more than $40,000 as band 3. 5. Save the data file if you are happy with the results of this exercise. Exercise 7: Computing variables Open the data file results.sav created in Exercise 2. This shows the exam scores for a class of high- school students. 1. Each student has a percentage mark for Maths, English, and History. Compute a new variable, named total, to calculate their total score out of 300. 2. We wish to compute the average (mean) mark over the three tests for each student. Compute a new variable, average, calculating this information. 3. Save the file. Using SPSS: Tasks 3 UCL Information Systems Exercise 8: Creating and saving output This exercise uses the file results.sav created in Exercise 2. 1. Use the Case Summaries command to produce a listing of your data. 2. Save the SPSS output in a file called result1.spo. 3. Produce a printout of the SPSS Viewer window. (Or use Print preview if you don't have access to a printer.) Exercise 9: Frequencies command This exercise uses the medicaltrialX data, and the incband variable as calculated earlier. If you do not have this, you can load the data file medicaltrialX-part2.sav. 1. Create a Frequencies command with a bar chart to find the following: The numbers within each of the three incband classifications. How many subjects are male, and how many female. 2. Create a Frequencies command for the two hormone level variables (i.e. hbefore and hafter). Include the following in the output: Do not display the actual frequency tables. A histogram with a normal curve superimposed for each variable. Produce values for the mean, mode and median. Objects from the SPSS Viewer can be copied and pasted into other applications (e.g. Microsoft Word or Excel). If you have time, open a new document in Microsoft Word. Copy and paste one of the histograms which you have just produced into the Word document. Save it with the title histogram.doc. Do not save the SPSS Viewer output. Exercise 10: Descriptives command Use the medicaltrialX data you have been working on — or you may wish to load the data file medicaltrialX-part2.sav. 1. Use the Descriptives command to display the default information for age and income. How old is the oldest participant? And the youngest? What is the average household income of the participants? 2. Produce a Descriptives command for the variables hbefore and hafter. This time use the Options to include the skewness and range in the output. Which of the two measurements has the largest range of readings? What is the range? Use the Help menu to find out what skewness means. Using this information, which of the two measurements do you think is closest to being normally distributed? Explain why. Using SPSS: Tasks 4 UCL Information Systems Exercise 11: Crosstabulations command This exercise uses the incband column as defined earlier. If you do not have this variable in your data, you can load the data file medicaltrialX-part2.sav. 1. Run a Crosstabs command for the variables incband and gender, including the following information: Each cell of the table should list the observed values, the expected values, and the unstandardised residuals. Also run the chi-square test. 2. Run a second Crosstabs command for the same variables. This time do not run the chi-square test, but include the following within each cell of the table: The row, column and total percentages. Exercise 12: Means command Use the medicaltrialX data you have been working on — or you may wish to load the data file medicaltrialX-part2.sav. 1. Build a Means command where the dependent variables are the two sets of hormone level measurements (i.e. hbefore and hafter) and the independent variable is incband. Answer the following questions: Which group achieves the highest average hormone level, and is this before or after treatment? Which group has the lowest mean hormone level before the treatment? Which group shows the most varied hormone levels before the treatment? And after? (Hint: look at the standard deviations.) 2. Build a similar means command as the previous one. This time make gender the independent variable. Answer the following question: Which gender showed the largest increase in blood hormone levels, on average? 3. Build another Means command which will again analyse the results for each subject, but this time looking at the results for each gender within each incband group. (Hint: enter one category variable in the independent list and click on the Next button before inserting the second category variable.) Which group of men show the highest mean level of H before treatment? What is that mean level? In the highest income category, which gender shows the highest mean level of H after treatment? Exercise 13: T Tests Use the medicaltrialX data you have been working on - or you may wish to load the data file medicaltrialX-part2.sav. 1. Perform a T Test to show whether there is a significant difference in the inital hormone levels (hbefore) between men and women. Is there? 2. Perform a T Test across all the cases, to decide whether there is a significant difference in the mean hormone concentrations before and after the treatment. Using SPSS: Tasks 5 UCL Information Systems 3. Now use the Select cases function to select only the women in the study, and repeat step (2). What do you find? If you have time, repeat the test, selecting men instead of women. Using SPSS: Tasks 6 UCL Information Systems Exercise 14: Correlation Before we test for correlations, we need to calculate a new variable. Refer to the Computing new data section in the course workbook if you need to. Use the medicaltrialX data you have been working on — or you may wish to load the data file medicaltrialX-part2.sav. 1. We need to calculate the change in blood levels of hormone H using the two values we already have (the blood levels before and after). Create a new variable dh using the Compute function to calculate the "after" level minus the "before" level. Now to perform the correlations. 2. Measure the strength of association between income and age with a correlation coefficient and its significance. 3. Measure the strength of association between age and dh with a correlation coefficient and its significance. 4. Produce a scatter plot of dh against age (placing dh on the vertical axis), and: Include a title. Produce a fit line on the graph (Hint: Double-click on the chart to edit it, then click on one of the data points to select the data points, and from the menus choose Chart | Add chart element | Fit line at total). Exercise 15: Regression Use the medicaltrialX data you have been working on — or you may wish to load the data file medicaltrialX-part2.sav. Perform a linear regression analysis, analysing the dependence of dh upon age (i.e. the same variables as in the previous exercise), drawing some conclusions about the regression line produced. Exercise 16: Graphical plots Use the medicaltrialX data you have been working on — or you may wish to load the data file medicaltrialX-part2.sav. 1. Produce a pie chart showing the number of people that fall into each of the three income bands. Include a title in the output. Produce labels for the frequency of each income band and the percentage of the population. 1. Produce a clustered bar chart summarising the mean "before" and "after" measurements separately, for men and for women (hint: gender is the category variable). Using SPSS: Tasks 7 UCL Information Systems