Transcript Intro to SPSS
Introduction to SPSS
Data types and SPSS data entry and analysis 1
In this session
What does SPSS look like? Types of data (revision) Data Entry in SPSS Simple charts in SPSS Summary statistics Contingency tables and crosstabulations Scatterplots and correlations Tests of differences of means 2
SPSS/PASW
3
Aspects of SPSS
Menus - Analyse and Charts esp.
Spreadsheet view of data Rows are cases (people, respondents etc.) Columns are Variables Variable view of data Shows detail of each variable type 4
Questionnaire Data Coding
5
In SPSS
We change ticks etc. on a questionnaire into numbers One number for each variable for each case How we do this depends on the type of variable/data 6
Types of data
Nominal Ranked Scales/measures Mixed types Text answers (open ended questions) 7
Nominal (categorical)
order is arbitrary e.g. sex, country of birth, personality type, yes or no.
Use numeric in SPSS
and give value labels.
(e.g. 1=Female, 2=Male, 99=Missing) (e.g. 1=Yes, 2=No, 99=Missing) (e.g. 1=UK, 2=Ireland, 3=Pakistan, 4=India, 5=other, 99=Missing) 8
Ranks or Ordinal
in order, 1st, 2nd, 3rd etc.
e.g. status, social class
Use numeric in SPSS with value labels
E.g. 1=Working class, 2=Middle class, 3=Upper class E.g. Class of degree, 1=First, 2=Upper second, 3=Lower second, 4=Third, 5=Ordinary, 99=Missing 9
Measures, scales
1.
Interval - equal units e.g. IQ 2.
Ratio - equal units, zero on scale e.g. height, income, family size, age Makes sense to say one value is twice another
Use numeric (or comma, dot or scientific) in SPSS
E.g. family size, 1, 2, 3, 4 etc.
E.g. income per year, 25000, 14500, 18650 etc.
10
Mixed type
Categorised data Actually ranked, but used to identify categories or groups e.g. age groups = ratio data put into groups
Use numeric in SPSS and use value labels
.
E.g. Age group, 1= ‘Under 18’, 2=‘18-24’, 3=‘25 34 ’, 4=‘35-44’, 5=‘45-54’, 6=‘55 or greater’ 11
Text answers
E.g. answers to open-ended questions Either enter text as given (
Use String in SPSS
)
Or
Code or classify answers into one of a small number types. (
Use numeric/nominal in SPSS
) 12
Data Entry in SPSS
Video by Andy Field 13
Frequency counts
Used with categorical and ranked variables e.g. gender of students taking Health and Illness option
Sex of student
Valid Female Male Total Frequency 25 9 34 Percent 73.5 Valid Percent 73.5 26.5 26.5 100.0 100.0 Cumulative Percent 73.5 100.0 14
e.g. Number of GCSEs passed by students taking Health and Illness option
Number of GCSEs
Valid 0 1 2 3 4 5 6 7 8 9 13 14 Total Frequency 1 1 4 2 3 1 1 34 6 4 2 6 3 Percent 2.9 Valid Percent 2.9 2.9 11.8 2.9 11.8 17.6 11.8 5.9 17.6 8.8 5.9 8.8 2.9 2.9 100.0 17.6 11.8 5.9 17.6 8.8 5.9 8.8 2.9 2.9 100.0 Cumulative Percent 2.9 5.9 17.6 35.3 47.1 52.9 70.6 79.4 85.3 94.1 97.1 100.0 15
Central Tendency
Mean
= average value sum of all the values divided by the number of values
Mode
= the most frequent value in a distribution (N.B. it is possible to have 2 or more modes, e.g. bimodal distribution)
Median
= the half-way value, or the value that divides the ordered distribution in the middle The middle score when scores are ordered N.B. need to put values into order first 16
Dispersion and variability
Quartiles The three values that split the sorted data into four equal parts.
Second Quartile = median.
Lower quartile = median of lower half of the data Upper quartile = median of upper half of the data Need to order the individuals first One quarter of the individuals are in each inter quartile range 17
Used on Box Plot
Age of Health and Illness students
Upper quartile Lower quartile Statist i c s
Age N Mean Median Valid Missing 34 0 24.03 21.00
Median
18
Variance
Average deviation from the mean, squared
Score Mean Deviation Squared Deviation
1 2 2.6
2.6
-1.6
-0.6
2.56
0.36
3 3 4 2.6
2.6
2.6
0.4
0.4
1.4
Total 0.16
0.16
1.96
5.20
5.20 is the Sum of Squares This depends on number of individuals so we divide by n (5) Gives
1.04
which is the
variance
19
Standard Deviation
The variance has one problem: it is measured in units squared.
This isn ’t a very meaningful metric so we take the square root value.
This is the Standard Deviation 20
Using SPSS
‘Analyse>Descriptive>Explore’ menu.
Gives mean, median, SD, variance, min, max, range, skew and kurtosis.
Can also produce stem and leaf, and histogram.
21
Charts in SPSS
Use ‘Chart Builder’ from ‘Graph’ menu or the Legacy menu And/or double click chart to edit it.
E.g. double click to edit bars (e.g. to change from colour to fill pattern).
Do this in SPSS first before cut and paste to Word Label the chart (in SPSS or in Word) 22
Stem and leaf plots
e.g. age of students taking Health and Illness option good at showing distribution of data outliers range 23
Stem and leaf plots e.g.
Age St em- and- Leaf Pl ot Fr equency St em & Leaf 6. 00 1 . 999999 17. 00 2 . 00000000001111134 5. 00 2 . 55678 3. 00 3 . 123 1. 00 3 . 5 2. 00 Ext r emes ( >=36) St em wi dt h: 10 Each l eaf : 1 case( s)
24
Box Plot
Statist i c s
Age N Mean Median Valid Missing 34 0 24.03 21.00 25
Box Plot
Fill colour changed.
N.B. numbers refer to case numbers.
26
Histograms and bar charts
Length/height of bar indicates frequency 27
Histogram
Fill pattern suitable for black and white printing 28
Changing the bin size
Bin size made smaller to show more bars 29
Pie chart
angle of segment indicates proportion of the whole 30
Pie Chart
Shadow and one slice moved out for emphasis
Analysing relationships
Contingency tables or crosstabulations Compares
nominal/categorical
variables But can include ordinal variables N.B. table contains counts (= frequency data) One variable on horizontal axis One variable on vertical axis Row and column total counts known as marginals
Example
In the Health and Illness class, are women more likely to be under 21 than men?
Crosstabulations
e.g.
Use column and row percentages to look for relationships
SPSS output
Chi-square
²
Cross tabulations and Chi-square are tests that can be used to look for a relationship between two variables: When the variables are
categorical
so the
data are nominal
(or frequency).
For example, if we wanted to look at the relationship between gender and age.
There are several different types of Chi-square ( ²), we will be using the
2 x 2 Chi-square
2x2 Chi-square results in SPSS
Another example
The Bank employees data
Bank Employees Chi-Square tests
Chi-Square analysis on SPSS
http://www.youtube.com/watch?v=Ahs8jS5m JKk 4m15s http://www.youtube.com/watch?v=IRCzOD27 NQU From 6m:30s to 9m:50s http://www.youtube.com/watch?v=532QXt1P M Q&feature=plcp&context=C3ba91a4UDOEgs ToPDskJ-ABupdp-Yfvuf4j4fJGzV 12m30s
Low values in cells
Get SPSS to output expected values Look where these are <5 Consider recoding to combine cols or rows
Tabulating questionnaire responses
Categorical survey data often “collapsed” for purposes of data analysis
Original category
White British White Irish Other White Indian Pakistani Bangladeshi Chinese Black British Afro-Caribbean African 16 30 12 2
Frequency
284 7 13 40 32 33
Collapsed category
White South Asian Chinese Black
Frequency
304 105 16 44
An analysis on a sample of 2 (e.g. Black African) would not have been very meaningful!
Recoding variables
http://www.youtube.com/watch?v=uzQ_522F 2SM&feature=related Ignore t-test for now 6m11s http://www.youtube.com/watch?v=FUoYZ_f6 Lxc Uses old version of SPSS, no submenu now. 6m
Scatterplots and correlations
Looks for association between variables, e.g.
Population size and GDP crime and unemployment rates height and weight Both variables must be rank, interval or ratio (scale or ordinal in SPSS).
Thus cannot use variables like, gender, ethnicity, town of birth, occupation.
44
Scatterplots
e.g. age (in years) versus Number of GCSEs 45
Interpretation
As Y increases X increases Called correlation Regression line model in red 46
Correlation measures association not causation
The older the child the better s/he is at reading The less your income the greater the risk of schizophrenia Height correlates with weight But weight does not
cause
height Height is
one
of the causes of weight (also body shape, diet, fitness level etc.) Numbers of ice creams sold is correlated with the rate of drowning Ice creams do not
cause
drowning (nor
vice versa
) Third variable involved – people swim more and buy more ice creams when it ’s warm 47
Scatterplot in SPSS
Use Graph menu http://www.youtube.com/watch?v=74BjgPQvI Eg 8m34s http://www.youtube.com/watch?v=blfflA 34pQ&feature=related 4m04s http://www.youtube.com/watch?v=UVylQoG4 hZM 1m50s, ignore polynomial regression 48
Modifying the Scatterplot
http://www.youtube.com/watch?v=803YCYA2 AoQ&feature=related 4m04s http://www.youtube.com/watch?v=vPzvuMuV Xk8&feature=related 3m40s 49
If mixed data sets
Change point icon and/or colour to see different subsets.
Overall data may have no relationship but subsets might.
E.g. show male and female respondents.
Use Chart builder 50
Correlation
Correlation coefficient = measure of strength of relationship, e.g. Pearson ’s
r
varies from 0 to 1 with a plus or minus sign
Correlati ons
Number of GCSEs Number of GCSEs Age Pearson Correlation Sig. (2-tailed) N Pearson Correlation Sig. (2-tailed) N *. Correlation is significant at the 0.05 level (2-tailed). 1 34 -.415
* .015 34 Age -.415
* .015 34 1 34 51
Positive correlation
as x increases, y increases
r = 0.7
52
Negative correlation
as x increases, y decreases
r = -0.7
53
Strong correlation (i.e. close to 1)
r = 0.9
54
Weak correlation (i.e. close to 0)
r = 0.2
55
Interpretation cont.
r 2
is a measure of
degree of variation in one variable accounted for by variation in the other
. E.g. If
r
=0.7 then
r 2
=.49 i.e. just under half the variation is accounted for (rest accounted for by other factors).
If
r
=0.3 then
r 2
=0.09 so 91% of the variation is explained by other things.
56
Significance of r
SPSS reports if
r
is significant at α=0.05
N.B. this is dependent on sample size to a large extent.
Other things being equal, larger samples more likely to be significant.
Usually, size of r is more important than its significance
57
Pearson ’s r in SPSS
http://www.youtube.com/watch?v=loFLqZmvf zU 6m57s 58
Parametric and non-parametric
Some statistics rely on the variables being investigated following a normal distribution. – Called
Parametric statistics
Others can be used if variables are not distributed normally – called
Non-parametric statistics.
Pearson ’s
r
is a parametric statistic Kendal ’s tau and Spearman’s rho (rank correlation) are non-parametric.
59
Assessing normality
Produce histogram and normal plot 60
Use statistical test
SPSS provides two formal tests for normality : Kolmogorov-Smirnov (K-S) and Shapiro Wilks (S-W) But, there is debate about KS Extremely sensitive to departure from normality May erroneously imply parametric test not suitable – especially in small sample So, always use a histogram as well.
61
Often can use parametric tests
Parametric tests (e.g. Pearson ’s
r
) are robust to departures from normality Small, non-normal samples OK But use non-parametric if Data are skewed (questionnaire data often is) Data are bimodal 62
Spearmans ’s rho
http://www.youtube.com/watch?v=r_WQe2c ISU From 4.14 to 4.56
http://www.youtube.com/watch?v=POkFi5vKv I8&feature=fvwrel 6m16s 63
So far…
Looked at relationships between nominal variables Gender vs age group Looked at relationships between scale variables Height vs. Weight Now combine the two
Groups
vs a scale variable E.g. Gender vs income 64
Reminder – IV vs DV
IV =
independent variable
What makes a difference, causes effects, is responsible for differences.
DV =
dependent variable
What is affected by things, what is changed by the IV.
Gender vs income. Gender = IV, income = DV So we investigate the effect of gender on income 65
Example 1 Age group vs. no. of GCSEs
Using the Health and Illness class data Age group defines 2 groups Under 21 21 and over Just two groups Can use
independent samples t-test
Independent because the two groups consist of different people.
t-test compares the means of the 2 groups.
66
Difference of means
Do under 21s have more or fewer GCSEs than 21 and overs?
Number of GCSEs Age group Under 21 21 and over
Group Statisti c s
N 16 Mean 6.44 Std. Deviation Std. Error Mean 3.140 .785 18 4.28 2.906 .685 Means are different (6.44 & 4.28) but is that significant?
67
Number of GCSEs Equal variances assumed Equal variances not assumed No significant difference therefore assume equal variances Number of GCSEs
Independent Samples T e s t
Levene's Test for Equality of Variances F Sig. t df Equal variances assumed .164 .689 2.082 32 Equal variances not assumed 2.073 30.789
Independent Samples T e s t
Sig. (2-tailed) .045 .047 t-test for Equality of Means Mean Difference 2.160 2.160 Std. Error Difference 1.037 1.042 95% Confidence Interval of the Difference Lower .047 Upper 4.272 .034 4.285 Levene's Test for Equality of Variances t-test for Equality of Means F .164 Sig. .689 t 2.082 2.073 df 32 30.789 Sig. (2-tailed) .045 .047 Mean Means are Difference statistically Std. Error Difference 1.037 1.042 68 95% Confidence Interval of the Difference Lower Upper .047 .034 4.272 4.285
Parametric vs non-parametric
Just as in the case of correlations, there are both kinds of tests. Need to check if DV is normally distributed.
Do this visually Also use statistical tests 69
Tests for normality
Kolmogorov-Smirnov and Shapiro-Wilk If n>50 use KS If n≤50 use SW Null hypothesis is ‘data are normally distributed’.
So if
p<0.05
then data are significantly different from a normal distribution –
use non parametric tests
If
p≥0.05
then no significant difference –
parametric tests use
70
Checking normality
Produce histogram of DV Tick box to undertake statistical test Interpret results.
71
t-test
Identify your two groups.
Determine what values in the data indicate those two groups (e.g. 1=female, 2=male) Select Analyze:Compare Means:Independent samples t-test http://www.youtube.com/watch?v=_KHI3ScO 8sc 9m40s 72
Mann-Whitney U test
Use this when comparing two groups and the DV is not normally distributed http://www.youtube.com/watch?v=7iTvv3m9d _g 3m45s 73
Comparing 3 or more groups
ANOVA = Analysis of Variance Analyze: Compare Means: One-way ANOVA http://www.youtube.com/watch?v=wFq1b3QjI 1U 4m04s Useful to get table of means (descriptives) and means plots from ANOVA options.
74
ANOVA Means and F value
75
ANOVA Means Plot
76