First Syntax

COMPUTE LabVot9209 =LabPoll09-LabVote92.
EXECUTE.

This article on the Ipsos Mori website by Dr Roger Mortimore has data from polls conducted during the 1992 general election and in 2009. On the right of the data there is a variable of Conservative vote change 92-09 but there is no Labour equivalent so I thought I would have a go at that. Turns out it was rather easy in the end. So easy I had time to make a nice graph as well.

If the graph is a little too small to see here on blogger click on the picture and it should expand to a much more reasonable size.

Steps to Syntax

Well the Collier book Using SPSS Syntax has arrived! Now as it's the last weekend before essay deadline I can't indulge it for long but there are a couple of things that I must show you. They're useful to some extent even if you don't want to go further down the syntax route so no excuses for not reading or doing.

• The Journal

Find or make a folder you want to have all your SPSS inputs sent to. I have mine as a sub folder of my course folder but you can have it where ever you like yours. Go Edit -> Options. On the first tab General the second box on the left entitled Session Journal check that the record syntax button is ticked and it's set to append just below then check the browse button just below to set your folder to recieve your SPSS inputs. Mine was previously set to the Temp folder which I guess must be the default option. This is a good thing to do even if you never use syntax as if you are unable to remember how you did an analysis and need to prove how you did it this provides a record which you can consult.

• The Log

If you have no interest in syntax this is a bit less relevant but if you're thinking about it for the future this is a great thing to do. Again under Edit -> Options but this time under the Viewer tab which is the one on the right of the General tab. Tick the box "Display commands in the log" which is on the bottom left and then click OK. This should in the output viewer display the syntax commands used to get your results.

Interaction in Factorial ANOVA

There is no interaction in graphs 1,5 and 6 while there is interaction in 2, 3 and 4. Of those that have interaction 2 and 4 have ordinal interaction where the graph lines remain separate but 3 has disordinal interaction as the graph lines cross each other. If there is interaction does this make the main effect pointless? Possibly but not necessarily. The most advanced analysis tool that can used in this situation is the Mk 1 Common Sense; not a lot of vendors stock that.

Now the sharper ones amongst you will notice that these graphs aren't done in SPSS in fact they are done in OpenOffice. It seemed a good idea at the time after all it's just the direction of the lines which is important; anyway this reminds me of a joke. Two accountants are sitting next to each other on a train. One leans over to the other and says "I see your not using Excel" and the other goes "No I'm using an alternative spreadsheet, I like to express my wild side"

Scripting conclusions

I've been involved in examining some data returned from a survey. The precise survey isn't important but I wanted to make a note of some good points for future reference when it comes to scripting.

If your doing a large scale survey running to thousands of respondents make sure you so a test run of the survey to pick up any design faults. 100 - 200 responses should be more than adequate. The question which you may think is perfectly logical in your head may in the mind of the public be the perfect prompt to give an answer which is pretty unhelpful.

For instance if you want to find out how often people visit something like the local park don't leave this as an open question. To do so is to create an open goal for respondents to give answers like usually, sometimes or my personal favourite regularly which could mean anything from hourly trips to an annual commemorative visit. Much better would be to use closed answers.

Respondents seemed happy to give their age and gender even if they didn't want to give their name. Number your scripts as you get them back. If your dealing with sensitive subjects it's best not to number them before you give them out as people are less likely to give truthful answers.

Don't ask people what their nationality of origin is. Many 2nd and 3rd generation members of the ethnic minorities will put British or British born which is fine but probably not the answer you want if your trying to find out people's ethnic origins.

Give plenty of options with the most common first particularly if the survey is to be carried out in an area with a wide variety of ethnicity's. There may be important differences between different groups which your study will benefit from recognising. For instance your not Sarah Palin; Africa is not a country but a continent of 53 countries. Recognise this and give a space for an open answer as well as tick boxes for the main ethnic groups you are likely to get given the location of your survey. Also give options for mixed race that go further one box so entitled.

Earlier versions of SPSS don't have the little pop up guide to the variable as your inputting the data. This will cause you pain as you dig around variable view which wont copy and paste and you have to type it all out to the windows notepad. Avoid this shame and upgrade.

Spend more time going through answers before you lump all and sundry into others otherwise others can get pretty big. It may be the best course of action to add further answers to your variable that you hadn't anticipated before you got the survey back but do this before you start inputting data.

The perils of the Amazon

Snakes, crocodiles and flesh eating bugs aren't yet available on Amazon.co.uk but just about everything else is. This includes Using SPSS Syntax: A Beginner's Guide by Jacqueline Collier and today I invested, note the positive synonym for spent, some of my hard earned on this. I should recieve it around the end of the week so expect more interesting syntax stuff to appear about a week after that as I have a deadline on the following Friday (19th).

Obama's launchpad

This is data from the 2008 General Social Survey. It shows identification with political parties amongst the US population. As far as pie charts go it does a job but I want to show some of the things the SPSS chart editor can do.

This is selected by double clicking on the object. First I think knowing the actual percentages would be good. This is found under Elements and then click on Show Data Labels. The numbers are a little small and these can be adjusted with the Text Style tab with Preferred Size. Changing the Font to Times New Roman with Bold gives it an extra punch. But there is a problem.

I think the data is too precise. Presenting it to 2 decimal places is over kill and doesn't add to understanding. To adjust this click the Number Format tab and set the number of decimal places to zero and the graph will present whole numbers. Next the colours could be made more representative. Double click on the coloured squares in the legend (the column on the right) to bring up the properties screen and then use the fill and border tab.

This graph really needs a proper title and this can be added with the add title button on the control panel this will bring up the properties panel with which you can adjust to your satisfaction the text but there is another problem the title is covering part of the legend. To fix this double click the title > Text Layout then use the Justify tool. I moved it left.

It's also imprtant to properly source things so we need to add a footnote to do this click on Options > Add Footnote then copy and paste the web address into the box. So as you can see it's starting to look better already.

But that is not all we can do. In US Presidential politics other party ID is not relevant so lets eliminate that. Go Properties then Categories then click on Other Parties and press the red cross button then apply.

Now Obama's victory was in many people's view won amongst the independents so we can highlight them by exploding a slice. First select independents on the legend, then Elements then explode slice. Don't select the whole pie as all the slices will be exploded if you do this.

An alternative explanation could be that the democratic base was more fired up after 8 years in opposition. Perhaps this was reflected in the Obama campaign's focus on community organisation which would have benefited most from having a large number of highly committed activists taking the Democrat message right to peoples front doors. So in the next version we look only at the proportions of strong party identifiers. This is again done with the categories tab. Next to which is the depth and angle tab. The effect is 3D. The angle is adjusted. I thought having the bigger Democratic slice first would be better so the order of slice was set to counterclockwise and finally distance was set to 1.

So there we have it a clear visual which shows that by 2008 the Democratic base was considerably larger than its Republican equivalent. Another startling success of the Presidency of George W. Bush.

First Post

The prime purpose of this blog is to get me to extend my SPSS skills and be able to demonstrate this to a wider public. I'm sure a wider public can't wait to see me scripting things on SPSS ... wait hello... COME BACK! Seriously this is going to be cool.

I'll also be looking at survey methods and reseach design as part of the MSc in Social Research Methods i'm doing. As a political junkie on the side I might stray onto the methodology of opinion polling. I also like pretty graphs. Pretty graphs can help change the world.