SPSS Training Manual

8/10/2019 SPSS Training Manual

1/50

SPSS

At the end of this session, you should be able to perform the following tasks

Introduction

The Statistical Package for the Social Sciences (SPSS) is a package of

programs for manipulating, analyzing, and presenting data; the package is

widely used in the social and eha!ioral sciences" The core program is called

SPSS #ase and there are a numer of add$on modules that e%tend the range

of data entry, statistical, or reporting capailities" &or e%ample, SPSS

'd!anced odels, SPSS egression odels, SPSS *%act Tests, SPSS Tales,

SPSS +ategories, SPSS Trends, SPSS issing alue 'nalysis etc"

Introduction to SPSS environment (Getting to know SPSS)

Starting SPSS

There are a numer of different ways to start SPSS-

. The simplest way is to look for an SPSS icon on your desktop" Place your cursor

on the icon and doule$click"

. /ou can also start SPSS y clicking on Start, mo!e your cursor up to Programs,

and then across to the list of programs a!ailale" o!e up or down until you find

SPSS for 0indows"

. SPSS will also start up if you doule$click on an SPSS data file listed in 0indows

*%plorer1these files ha!e a "sa! e%tension"

0hen you open SPSS you may encounter this grey front co!er screen asking 20hat

would you like to do34"


2/50

5t is easier to close this screen (click on the cross in the top right$hand corner) and

get used to using the other SPSS menus" 0hen you close the opening screen you

will see a lank spreadsheet" To open an e%isting SPSS data file from this


3/50

spreadsheet screen, click on File, and then Open, from the menu displayed at the

top of the screen"

pening an e!isting data "le

5f you wish to open an e%isting data file, click on &ile from the menu across the top

of the screen, and then choose 6pen, and then 7ata" The 6pen &ile dialogue o%

will allow you to search through the !arious directories on your computer to find

where your data file is stored" /ou should always open data files from the hard

dri!e of your computer, not from e%ternal memory dri!e"

Saving a data "les

0hen you first create a data file, or make changes to an e%isting one, you must

rememer to sa!e your data file" This does not happen automatically, as in some

word processing programs" 5f you don4t sa!e regularly, and there is a power

lackout or you accidentally press the wrong key, you will lose all of your work" So

sa!e yourself the heartache and sa!e regularly" To sa!e a file you are working on,

go to the File menu(top left$hand corner) and choose Save"

0hen you first sa!e a new data file, you will e asked to specify a name for the file

and to indicate a directory and a folder that it will e stored in" +hoose the directory

and then type in a file name" SPSS will automatically gi!e all data file names the

e%tension .sav" This is so that it can recognise it as an SPSS data file" 7on4t change

this e%tension, otherwise SPSS won4t e ale to find the file when you ask for it

again later"

Starting a new data file

Starting a new data file is easy in SPSS is easy" +lick on &ile, then, from the

dropdown menu, click on 8ew and then 7ata" &rom here you can start defining your

!ariales and entering your data"

SPSS windows

The main windows you will use in SPSS are the 7ata *ditor, the iewer, the Pi!ot

Tale *ditor, +hart *ditor and the Synta% *ditor"


4/50

Data Editor window

The 7ata *ditor window displays the contents of your data file, and in this window

you can open, sa!e and close e%isting data files; create a new data file; enter data;

make changes to the e%isting data file; and run statistical analyses, the data editorwindow also contains the !ariale !iew

Viewer window

iewer window will open automatically when you open the data editor !iew,

This

window displays the results of the analyses you ha!e conducted, including tales

and charts" 5n this window you can modify the output, delete it, copy it, sa!e it, or

e!en transfer it into a 0ord document" 0hen you sa!e the output from SPSS

statistical analyses it is sa!ed in a separate file with a "spo e%tension, to distinguish

it from data files, which ha!e a "sa! e%tension"

The iewer screen consists of two parts" 6n the left is an outline or menu pane,

which gi!es you a full list of all the analyses you ha!e conducted" /ou can use this

side to 9uickly na!igate your way around your output (which can ecome !ery long,

!ery 9uickly)" :ust click on the section you want to mo!e to and it will appear on

the right$hand side of the screen" 6n the right$hand side of the iewer window are

the results of your analyses, which can include tales and charts (or graphs)"

Saving output


5/50

To sa!e the results of your analyses you must ha!e the iewer window open on the

screen in front of you" +lick on &ile from the menu at the top of the screen" +lick on

Sa!e" +hoose the directory and folder you wish to sa!e your output in, and then

type in a file name that uni9uely identifies your output" +lick on Save"

NoteTo pre!ent confusion and frustration, get organized and keep good records of the

analyses you ha!e done and of where you ha!e sa!ed the results"

Printing output

/ou can use the menu pane (left$hand side) of the iewer window to select

particular sections of your results to print out" To do this you need to highlight the

sections that you want" +lick on the first section you want, hold the +trl key on your

keyoard down and then ust click on any other sections you want" To print these

sections, click on the &ile menu (from the top of your screen) and choose Print"

SPSS will ask whether you want to print your selected output or the whole output"

Pivot Table Editor window

The tales you see in the iewer window (which SPSS calls Pi!ot Tales) can emodified to suit your needs" To modify a tale you need to doule$click on it, which

takes you into what is known as the Pi!ot Tale *ditor" /ou can use this editor to

change the look of your tale, the size, the fonts used, the dimensions of the

columns1you can e!en swap the presentation of !ariales around from rows to

columns"


6/50

Chart Editor window

0hen you ask SPSS to produce a histogram, ar graph or scatterplot, it initially

displays these in the iewer window" 5f you wish to make changes to the type or

presentation of the chart, you need to go into the +hart *ditor window y doule$clicking on your chart" 5n this window you can modify the appearance and format of

your graph, change the fonts, colours, patterns and line markers"

Snta! Editor window

The Synta% *ditor is particularly useful when you need to repeat a lot of analyses or

generate a numer of similar graphs" /ou can use the normal SPSS menus to set up

the asic commands of a particular statistical techni9ue and then 2paste4 these to

the Synta% *ditor (see &igure


7/50

This is done in a dialogue o%" &or e%ample, when you ask SPSS to run &re9uencies

it will display a dialogue o% asking you to nominate which !ariales you want to

use"

&rom here, you will open a numer of additional su$dialogue o%es, where you will

e ale to specify which statistics you would like displayed, the charts that you

would like generated and the format the results will e presented in" 7ifferent

options are a!ailale, depending on the procedure or analysis to e performed, ut

the asic principles in using dialogues o%es are the same"

Sele"ting variables in a dialogue bo!

To indicate which !ariales you want to use, you need to highlight the selected

!ariales in the list pro!ided (y clicking on them), then click on the arrow utton

to mo!e them into the empty o% laeled ariale(s)" To remo!e a !ariale from

the o%, you ust re!erse the process" +lick on the !ariale in the ariale(s) o%

that you wish to remo!e, click on the arrow utton, and it shifts the !ariale ack

into the original list" /ou will notice the direction of the arrow utton changes,

depending on whether you are mo!ing !ariales into or out of the ariale(s) o%"

Dialogue bo! buttons

5n most dialogue o%es you will notice a numer of standard uttons (6>, Paste,

eset, +ancel and ?elp; see &igure


8/50

. 6>- click on this utton when you ha!e selected your !ariales and are ready to

run the analysis or procedure"

. Paste- this utton is used to transfer the commands that SPSS has generated in

this dialogue o% to the Synta% *ditor (a description of which is presented earlier inthis chapter)" This is useful if you wish to repeat an analysis a numer of times, or

if you wish to make changes to the SPSS commands"

. eset- this utton is used to clear the dialogue o% of all the pre!ious commands

you might ha!e gi!en when you last used this particular statistical techni9ue or

procedure" 5t gi!es you a clean slate to perform a new analysis, with different

!ariales"

+ancel- clicking on this utton closes the dialogue o% and cancels all of the

commands you may ha!e gi!en in relation to that techni9ue or procedure"

. ?elp- click on this utton to otain information aout the techni9ue or procedure

you are aout to perform"

%losing SPSS

0hen you ha!e finished your SPSS session and wish to close the program down,

click on the &ile menu at the top left of the screen" +lick on *%it" SPSS will promptyou to sa!e your data file and a file that contains your output (results of the

analyses)" SPSS gi!es each file an e%tension to indicate the type of information that

it contains" ' data file will e gi!en a "sa! e%tension, while the output files will e

assigned a "spo e%tension"

%reating a data "le and entering data


9/50

The asic steps in!ol!ed in the process of creating a data file and entering

the data using SPSS

A +heck and modify, where necessary, the options (or preferences, as

they were referred to in earlier !ersions of SPSS) that SPSS uses to

display the data and the output that is produced"

B Set up the structure of the data file y 2defining4 the !ariales"< *nter the data"

%hanging the SPSS &ptions'

The options allow you to define how your !ariales will e displayed, the size of

your charts, the type of tales that will e displayed in the output and many other

aspects of the program"

To open the 6ptions screen, click on Editfrom the menu at the top of the screen

and then choose Options"


10/50

Tips

#d$ust as it suit ou. E!periment with different options at our own

"onvenien"e

$e"ning the variables

7efining the !ariales means informing SPSS aout your !ariale names and coding

instructions" /ou will do this in the Data Editor window, the 7ata *ditor window

consists of two different !iews- Data Viewand Variable View

Data Editor window

Variable View

/our o now is to define each of your !ariales y specifying the re9uired

information for each !ariale listed in your codeook" Some of the information you

will need to pro!ide yourself (e"g" name); other its are pro!ided automatically y

SPSS using default !alues" These default !alues can e changed if necessary"Values

5n the alues column you can define the meaning of the !alues you ha!e used to

code your !ariales" 5 will demonstrate this process for the !ariale 2Se%4"

A" +lick on the three dots on the right$hand side of the cell" This opens the alue

Cael dialogue o%"


11/50

B" +lick in the o% marked alue" Type in A"


12/50

Danger alert&&&

'hen entering data( remember to save our data file regularl. SPSS does

not automati"all save it for ou. )f ou don*t save it( ou ris+ losing all the

information ou have entered. To save( $ust "li"+ on the File menu and

"hoose Save or "li"+ on the i"on that loo+s li+e a "omputer dis+.

#odifying the data "le

'fter you ha!e created a data file you may need to make changes to it (e"g" to add,delete or mo!e !ariales; or to add or delete cases)" There are also situations

where you may need to sort a data file into a specific order, or to split your file to

analyse groups separately"

To delete a "ase

To insert a "ase between e!isting "ases

To delete a variable

To insert a variable between e!isting variables

To move an e!isting variable

To sort the data file


13/50

/ou can ask SPSS to sort your data file according to !alues on one of your !ariales

(e"g" se%, age)" +lick on the 7ata menu, choose Sort +ases and specify which

!ariale will e used to sort y" To return your file to its original order, repeat the

process, asking SPSS to sort the file y 57"

To split the data file

Sometimes it is necessary to split your file and to repeat analyses for groups (e"g"

males and females) separately" Please note that this procedure does not physically

alter your file in any permanent manner" 5t is an option you can turn on and off as it

suits your purposes" The order in which the cases are displayed in the data file will

change, howe!er" /ou can return the data file to its original order (y 57) y using

the Sort +ases command descried ao!e"

Steps

A" ake sure you ha!e the 7ata *ditor window open on the screen"

B" +lick on the 7ata menu and choose the Split &ile option"


14/50

&or some analyses you may wish to select a suset of your sample (e"g" only

males)"


B" +lick on the 7ata menu and choose the Select +ases option"

"

&or the analyses (e"g" correlation) that you perform after this select cases

procedure, only the group that you selected (e"g" males) will e included" 0hen you

ha!e finished the analyses, you need to go ack and turn the Select +ases option

off"

To turn the sele"t "ases option off


B" +lick on the 7ata menu and choose Select +ases option"


15/50

7ata files can e prepared in the icrosoft *%cel program and then imported into

SPSS for analysis" 6ne word of warning- *%cel can cope with only B@E columns of

data (in SPSS language- !ariales)" 5f your data file is likely to e larger than this,

it is proaly easier to set it up in SPSS, rather con!ert from *%cel to SPSS later"

Step A- Set up the !ariale names

Note,The !ariale names must conform to the SPSS rules for naming !ariales

Step B- *nter the data

Step " The data will appear

on the screen with the !ariale names listed across the top" /ou will, howe!er, need

to go ahead and define the ariale laels, alue laels and the type of easure"

Step -, Saving as an SPSS file

0hen you ha!e completed this process of fully defining the !ariales, you need to

sa!e your file as an SPSS file" +hoose &ile, and then Sa!e 's from the menu at the

top of the screen" Type in a suitale file name" ake sure that the Sa!e as Type is

set at SPSS (H"sa!)" +lick on Sa!e" 0hen you wish to open this file later to analyse

your data using SPSS, make sure you choose the file that has a "sa! e%tension (not

your original *%cel file that has an "%ls e%tension)"

Screening and cleaning the data

#efore you start to analyse your data it is essential that you check your data set for

errors" 5t is !ery easy to make mistakes when entering data, and unfortunately


16/50

some errors can completely mess up your analyses" &or e%ample, entering


17/50

Step +* inding the error in the data "le

Pro"edures for identifing the "ase where an error has o""urred

ethod A

A" ake sure that the 7ata *ditor window is open and on the screen in front of you

with the data showing"

B" +lick on the !ariale name of the !ariale in which the error has occurred (e"g"

se%)"


18/50

=" 5n the Cael cases section choose 57 from your !ariale list" This will gi!e you the

57 numer of the case, and will allow you to trace ack to the 9uestionnaireIrecord

with the mistake"

@" 5n the Statistics section choose 6utliers" To sa!e unnecessary output you mayalso like to remo!e the tick from 7escripti!es (ust click once)" +lick on +ontinue"

E" 5n the 6ptions section choose *%clude cases pairwise" +lick on +ontinue and then

6>"

Step -* %orrecting the error in the data "le

Pro"edure for "orre"ting the error in the data file

A" To correct the error, it will e necessary to go ack to your 9uestionnaires (or the

records from your e%periment)" &ind the 9uestionnaire or record with the 57

numer that was identified as an e%treme !alue" +heck what !alue should ha!e

een entered for that person (e"g" for se%- was it a male (score A) or a female

(score B)3)"

B" o!e down to the case numer that has the error" 6nce you ha!e found the

person with that case numer, mo!e across the row until you come to the column of

the !ariale with the error (e"g" Se%)" Place the cursor in the cell, make sure that itis highlighted and then ust type in the correct !alue"

This will replace the old incorrect !alue" Press one of the arrow keys and you will

see the correct !alue appear in the cell"

Note, 'fter you ha!e corrected your errors it is a good idea to repeat &re9uencies

to doule$check" Sometimes, in correcting one error, you will ha!e accidentally

caused another error" 'lthough this process is tedious it is !ery important that you

start with a clean, error$free data set" The success of your research depends on itJ

7on4t cut corners"

$escriptive statistics

7escripti!e statistics ha!e a numer of uses" These include-


19/50

. To descrie the characteristics of your sample in the ethod section of your

report;

. To check your !ariales for any !iolation of the assumptions underlying the

statistical techni9ues that you will use to address your research 9uestions;and

. To address specific research 9uestions"

%ategorical variables

To otain descripti!e statistics for categorical !ariales you should use &re9uencies"

Procedure for otaining descripti!e statistics for categorical !ariales

A" &rom menu, #nale, then click on Des"riptive Statisti"s, then

Fre/uen"ies"

B" +hoose and highlight the categorical !ariales you are interested in (e"g"

se%)"o!e these into the ariales o%"


20/50

Note on s+ewness and +urtosis

5nformation on skewness and kurtosis may e needed if these !ariales are to e

used in parametric statistical techni9ues (e"g" t$tests, analysis of !ariance)" The

skewness !alue pro!ides an indication of the symmetry of the distriution" >urtosis,on the other hand, pro!ides information aout the 2peakedness4 of the distriution"

5f the distriution is perfectly normal, you would otain a skewness and kurtosis

!alue of K (rather an uncommon occurrence in the social sciences)" Positi!e

skewness !alues indicate positi!e skew (scores clustered to the left at the low

!alues)" 8egati!e skewness !alues indicate a clustering of scores at the high end

(right$hand side of a graph)" Positi!e kurtosis !alues indicate that the distriution is

rather peaked (clustered in the centre), with long thin tails" >urtosis !alues elow K

indicate a distriution that is relati!ely flat (too many cases in the e%tremes)" 0ith

reasonaly large samples, skewness will not 2make a sustanti!e difference in the

analysis4 (Taachnick L &idell, BKKA, p" F=)" >urtosis can result in an underestimate

of the !ariance, ut this risk is also reduced with a large sample (BKKM cases- see

Taachnick L &idell, BKKA, p" F@)"

.sing graphs to describe and e!plore the data

/istograms

?istograms are used to display the distriution of a single continuous !ariale (e"g"

age, percei!ed stress scores)"

Pro"edure for "reating a histogram

A" &rom the menu at the top of the screen click on- Nraphs, then click on

?istogram"

B" +lick on your !ariale of interest and mo!e it into the ariale o%" This

should e a continuous !ariale (e"g" total percei!ed stress)"


21/50

=" 5f you wish to gi!e your graph a title click on the Titles utton and type the

desired title in the o% (e"g" ?istogram of Percei!ed Stress scores)"

@" +lick on Continue, and then O0"

0ar graphs

#ar graphs can e simple or !ery comple%, depending on how many !ariales you

wish to include"

Procedure for creating a ar graph

A" &rom the menu at the top of the screen click on- Nraphs, then #ar"

B" +lick on +lustered"


22/50

pro!ides a general indication of the strength of the relationship etween your two

!ariales"

Pro"edure for "reating a s"atterplot

A" &rom the menu at the top of the screen click on- Nraphs, then on Scatter"

B" +lick on Simple and then 7efine"

"

#anipulating the data

D#N2E3 #4E3T

ake sure you ha!e a second copy of your data set" 5f you make a mistake here

you will lose or corrupt your original data" Therefore, it is essential that you ha!e a

ackup copy" ememer that, unlike other spreadsheet programs (e"g" *%cel),

SPSS does not automatically recalculate !alues if you add e%tra cases or if you

make changes to any of the !alues in the data file" Therefore you should perform


23/50

the procedures illustrated in this chapter only when you ha!e a complete (and

clean) data file

%alculating total scale scores

#efore you can perform statistical analyses on your data set you need to calculate

total scale scores for any scales used in your study" This in!ol!es two steps-

. Step A- re!ersing any negati!ely worded items

A" &rom the menu, click on- Transform, then click on ecode, then 5nto Same

ariales"

B" Select the items you want to re!erse (opB, op=, opE)" o!e these into the

ariales o%"


24/50

. Step B- instructing SPSS to add together scores from all the items that make up

the suscale or scale"

A" &rom the menu at the top of the screen click on- Transform, then click on

+ompute"

B" 5n the Target !ariale o% type in the new name you wish to gi!e to the total

scale scores (it is useful to use a T prefi% to indicate total scores as this makes

them easier to find in the alphaetical list of !ariales when you are doing your

analyses)"

)mportant. ake sure you do not accidentally use a !ariale name that has

already een used in the data set" 5f you do, you will lose all the original data1

potential disaster; so check your codeook"


25/50

details of what was done to "al"ulate it. )f an items were reversed( this

should be spe"ified along with details of whi"h items were added to "reate

the s"ore.

%ollapsing a continuous variable into groups

&or some analyses (e"g" 'nalysis of ariance) you may wish to di!ide the

sample into e9ual groups according to respondents4 scores on some !ariale

(e"g" to gi!e low, medium and high scoring groups)" isual #inning to

identify suitale cut$off points to reak the continuous !ariale age into

three appro%imately e9ual groups" The same techni9ue could e used to

create a 2median split4- that is, to di!ide the sample into two groups, using

the median as the cut$off point" 6nce the cut$off points are identified, isual#inning will create a new (additional) categorical !ariale that has only three

!alues, corresponding to the three age ranges chosen" This techni9ue lea!es

the original !ariale age, measured as a continuous !ariale, intact so that

you can use it for other analyses"

Procedure for collapsing a continuous !ariale into groups

A" Transform, and choose Visual 5inning"

B" Select the continuous !ariale that you want to use (e"g" age)" Transfer it

into the ariales to #and o%" +lick on the +ontinue utton"


26/50

E" +lick on the ake Caels utton ack in the main dialogue o%" This will

automatically generate !alue laels for each of the new groups created"/ou

can modify these if you wish y clicking in the cells of the Nrid"

F" +lick on 6> and a new !ariale ('geand


27/50

@" &or the second !alue 5 would type B in the 6ld alue, ut in the 8ew alue 5

would type A" This will recode all the !alues of oth A and B from the original coding

into one group in the new !ariale to e created with a !alue of A"

E" &or the third !alue of the original !ariale 5 would type A in the 6ld alue and Bin the 8ew alue" This is ust to keep the !alues in the new !ariale in se9uence"

+lick on 'dd" epeat for all the remaining !alues of the original !alues" 5n the tale

6ld8ew you should see the following codes for this e%ample- AA; BA;


28/50

'e will be using the surve data "alled 6776 8.S. 2eneral So"ial Surve in

the installed SPSS.

?ypothesis

?o- There is no significant relationship etween le!el of educational 9ualification

and 6ccupational Prestige

?A- There is a significant relationship etween le!el of educational 9ualification

and 6ccupational Prestige

Preliminar analses for "orrelation

Pro"edure for "al"ulating Pearson produ"t9moment "orrelation

A" &rom the menu at the top of the screen click on- 'nalyze, then click on

+orrelate, then on #i!ariate"

B" Select your two !ariales and mo!e them into the o% marked ariales (e"g"

total percei!ed stress, total P+65SS)"/ou can list a whole range of !ariales here,

not ust two" 5n the resulting matri%, the correlation etween all possile pairs of

!ariales will e listed" This can e 9uite large if you list more than ust a few

!ariales"


29/50

@" +lick 6>"

The output generated from this procedure is shown elow"

Highest Year of

School CompletedR's Occupational Prestige

Score (1980

Highest Year of School Completed Pearson Correlation 1 !"#0($$

Sig! (#%tailed !000

& 1"10 11"

Respondent's OccupationalPrestige Score (1980

Pearson Correlation!"#0($$ 1

Sig! (#%tailed !000

& 11" 118

$$ Correlation is significant at the 0!01 leel (#%tailed!

)nterpretation of output from "orrelation

A +hecking the information aout the sample (8)

B 7etermining the direction of the relationship ($!e or M!e)

< 7etermining the strength of the relationship (the !alue of r)

= +alculating the coefficient of determination

@ 'ssessing the significance le!el

Presenting the results from "orrelation

The results of the ao!e e%ample could e presented in a research report as

follows-

The relationship etween le!el of educational 9ualification (as measured y the

?ighest /ear of School +ompleted) and 6ccupational Prestige (as measured y the

espondentUs 6ccupational Prestige Score) was in!estigated using Pearson product$

moment correlation coefficient" There was a strong, negati!e correlation etween

the two !ariales VrD"@B, nDA=A@, pW"KKK@X, with high le!els of educational

9ualification associated with high le!els of 6ccupational Prestige"

Comparing the "orrelation "oeffi"ients for two groups


30/50

Sometimes when doing correlational research you may want to compare the

strength of the correlation coefficients for two separate groups" &or e%ample, you

may want to look at the relationship etween educational 9ualification and

6ccupational Prestige for males and females separately"

Pro"edure for "omparing "orrelation "oeffi"ients for two groups of sub$e"ts

'- Split the sample

A" ake sure you ha!e the 7ata *ditor window open on the screen in front of you"

(5f you currently ha!e the iewer 0indow open, click on 0indow and choose the

SPSS 7ata *ditor")

B" &rom the menu at the top of the screen click on- 7ata, then click on Split &ile"

"

@" This will split the sample y se% and repeat any analyses that follow for these

two groups separately"

#- +orrelation

A" &ollow the steps in the earlier section of this chapter to re9uest the correlation

etween your two !ariales of interest (e"g" educational 9ualification, 6ccupational

Prestige)"

These will e reported separately for the two groups"


espondentUs Se%

?ighest /ear ofSchool

+ompleted

Us 6ccupationalPrestige Score

(AGK)

ale ?ighest /ear of School+ompleted

Pearson +orrelationA "==(HH)

Sig" (B$tailed) "KKK 8 E


31/50

Us 6ccupational PrestigeScore (AGK)

Pearson +orrelation"==(HH) A

Sig" (B$tailed) "KKK

8 EA EBA

&emale ?ighest /ear of School+ompleted

Pearson +orrelationA "@=K(HH)


8 GFF FE

Us 6ccupational PrestigeScore (AGK)

Pearson +orrelation"@=K(HH) A


8 FE FF

$$ Correlation is significant at the 0!01 leel (#%tailed!

:Please remember, when you ha!e finished looking at males and females

separately you will need to turn the Split &ile option off" 5t stays in place until you

specifically turn it off" To do this, +lick on Data, Split Fileand click on the first

utton- #nale all "ases( do not "reate groups" +lick on O0"X

Partial "orrelation

Partial correlation is similar to Pearson product$moment correlation, e%cept that it

allows you to control for an additional !ariale" This is usually a !ariale that you

suspect might e influencing your two !ariales of interest" #y statistically

remo!ing the influence of this confounding !ariale, you can get a clearer and more

accurate indication of the relationship etween your two !ariales"

Details of e!ample

Procedure for partial correlation

A" &rom the menu at the top of the screen click on- 'nalyze, then click on

+orrelate, then on Partial"

B" +lick on the two continuous !ariales that you want to correlate (e"g" ?ighest

/ear of School +ompleted, 6ccupational Prestige score)" +lick on the arrow to mo!e

these into the ariales o%"


32/50

"

+ontrol ariales

?ighest/ear ofSchool

+ompleted

Us6ccupationa

l PrestigeScore(AGK)

'ge ofespondent

$none$(a) ?ighest /ear ofSchool +ompleted

+orrelationA"KKK "@BK $"B@=

Significance (B$tailed) " "KKK "KKK df K A=A< A@KE Us 6ccupational

Prestige Score(AGK)

+orrelation"@BK A"KKK "KKF

Significance (B$tailed) "KKK " "F df A=A< K A=A=

'ge ofespondent +orrelation $"B@= "KKF A"KKK

Significance (B$tailed) "KKK "F " df A@KE A=A= K'ge ofespondent

?ighest /ear ofSchool +ompleted

+orrelationA"KKK "@


33/50

+ompleted and 6ccupational Prestige), not controlling for your other !ariale" 5n

this case the correlation is "@BK" The word 2none4 in the left$hand column indicates

that no control !ariale is in operation"

B" The ottom half of the tale repeats the same set of correlation analyses, utthis time controlling for (taking out) the effects of your control !ariale (e"g" age)"

5n this case the new partial correlation is "@


34/50

Some of the main types of research 9uestions that multiple regression can e used

to address are-

. ?ow well a set of !ariales is ale to predict a particular outcome;

. 0hich !ariale in a set of !ariales is the est predictor of an outcome; and

. 0hether a particular predictor !ariale is still ale to predict an outcome when the

effects of another !ariale are controlled for

%a$or tpes of multiple regression

A Standard multiple regression

B ?ierarchical multiple regression

< Stepwise multiple regression

#ssumptions of multiple regression

Details of e!ample

To illustrate the use of multiple regression we will e%amine the effect of 'ge of

espondent, ?ighest /ear of School +ompleted, ?ighest /ear School +ompleted$

father, ?ighest /ear School +ompleted$ other, ?ighest /ear School +ompleted$

Spouse on espondentUs 6ccupational Prestige"

?o- 'ge of espondent, ?ighest /ear of School +ompleted, ?ighest /ear School

+ompleted$ father, ?ighest /ear School +ompleted$ other and ?ighest /ear

School +ompleted$ Spouse will not significantly predict espondentUs

6ccupational Prestige"

?o- 'ge of espondent, ?ighest /ear of School +ompleted, ?ighest /ear School

+ompleted$ father, ?ighest /ear School +ompleted$ other and ?ighest /ear

School +ompleted$ Spouse will significantly predict espondentUs

6ccupational Prestige"

To e%plore this hypothesis, we will e using standard multiple regression

Standard multiple regression


35/50

Procedure for standard multiple regression

A" 'nalyze, then click on egression, then on Cinear"

B" +lick on your continuous dependent !ariale (espondentUs 6ccupational Prestige

score and mo!e it into the 7ependent o%"


36/50

(percei!ed stress) is e%plained y the model (which includes the !ariales of Total

astery and Total P+65SS)" 5n this case the !alue is "=EG" *%pressed as a

percentage (multiply y AKK, y shifting the decimal point two places to the right),

this means that our model (which includes astery and P+65SS) e%plains =E"G per

cent of the !ariance in percei!ed stress" /ou will notice that SPSS also pro!ides an

'dusted S9uare !alue in the output" 0hen a small sample is in!ol!ed, the

s9uare !alue in the sample tends to e a rather optimistic o!erestimation of the

true !alue in the population (see Taachnick L &idell, BKKA, p" A=F)" The 'dusted

s9uare statistic 2corrects4 this !alue to pro!ide a etter estimate of the true

population !alue" 5f you ha!e a small sample you may wish to consider reporting

this !alue, rather than the normal S9uare !alue"

To assess the statistical significance of the result it is necessary to look in the talelaelled '86'" This tests the null hypothesis that multiple in the population

e9uals K" The model in this e%ample reaches statistical significance (Sig D "KKK, this

really means pW"KKK@)"

ANOVA(b)

)odelSum ofS*uares df )ean S*uare Sig!

1 Regression #95.#!89# " "8.!".8 8!86 !000(a

Residual 6966.!5. "." 1#1!161

7otal 9900!5#9 "80

a Predictors/ (Constant Highest Year School Completed Spouse +ge of Respondent Highest Year SchoolCompleted ather Highest Year of School Completed Highest Year School Completed )other2 3ependent 4aria2le/ R's Occupational Prestige Score (1980

)odel

nstandardied Coefficients

Standardied

Coefficients t Sig!

9": Confidence;nteral for < Correlations

Collinearit=Statistics

"


This tale gi!es oth etween$groups and within$groups sums of s9uares, degrees

of freedom etc" /ou will recognise these from your statistics ooks" The main thing

you are interested in is the column marked Sig" 5f the Sig" !alue is less than or

e9ual to "K@ (e"g" "K


47/50

The results of the one$way etween$groups analysis of !ariance with post$hoc tests

could e presented as follows-

' one$way etween$groups analysis of !ariance was conducted to e%plore the

impact of age on le!els of optimism, as measured y the Cife 6rientation test(C6T)" Suects were di!ided into three groups according to their age (Nroup A- B

or less; Nroup B-


48/50

two !ariales- one independent !ariale (categorical) (e"g" Time AI Time BI Time


49/50

%ultivariate tests

5n this tale the !alue that you are interested in is 'il+s* 4ambda, also the

associated proaility !alue gi!en in the column laelled Sig. 'll of the multi!ariate

tests yield the same result, ut the most commonly reported statistic is 0ilks4Camda" 5n this e%ample the !alue for 0ilks4 Camda is "B@, with a proaility !alue

of "KKK (which really means pW"KKK@)" The p !alue is less than "K@; therefore we

can conclude that there is a statistically significant effect for time" This suggests

that there was a change in confidence scores across the three different time

periods"

Presenting the results from one9wa repeated measures #NOV#

The results of a one$way repeated measures '86' could e presented as follows-

' one$way repeated measures '86' was conducted to compare scores on the

+onfidence in +oping with Statistics test at Time A (prior to the inter!ention), Time

B (following the inter!ention) and Time < (three$month follow$up)" The means and

standard de!iations are presented in Tale OO" There was a significant effect for

time V0ilks4 CamdaD"B@, &(B, BG)D=A"AF, pW"KKK@"

Two9wa between9groups #NOV#

Details of e!ample

)nterpretation of output from two9wa #NOV#

Presenting the results from two9wa #NOV##dditional analses if ou obtain a signifi"ant intera"tion effe"t

%i!ed between9within sub$e"ts analsis of varian"e

Details of e!ample

)nterpretation of output from mi!ed between9within #NOV#

Presenting the results from mi!ed between9within #NOV#

Non9parametri" statisti"s

Chi9s/uare

%ann9'hitne 8 Test

'il"o!on Signed 3an+ Test0rus+al9'allis Test


50/50

Friedman Test

Spearman*s 3an+ Order Correlation

SPSS Training Manual

Documents