SPSS MADE SIMPLE: 2.0

INTRODUCTION to EXPERIMENTAL DESIGNS & STAT TESTS: SPSS MADE SIMPLE: 2.0

PUT SIMPLY: this blog post is more for...

..."PSYCHOLOGY STUDENTS". 

• Before you can even input any DATA from an experimental design... you MUST FIRST establish the TYPE of EXPERIMENTAL DESIGN. 

• This MUST BE DONE in order to understand HOW to input your data within any STAT PACKAGE e.g. SPSS.

• WHY BOTHER to know the BASICS of EXPERIMENTAL DESIGNS within SPSS?

• ANSWER: TO understand if a study's findings rings any truth. Just because it may use brain scans etc does not make it valid! 

• There are TWO KEY TYPES of EXPERIMENTAL (EXP) DESIGNS, confusingly enough though it has a number of names for the same type! 

• REPEATED MEASURES is also known as MATCH PAIR SAMPLE or WITHIN SUBJECTS DESIGN! 

• INDEPENDENT MEASURES is also known as BETWEEN SUBJECT DESIGN or INDEPENDENT SAMPLE. 

• WITHIN SUBJECTS DESIGN is called this because the study focuses on the subjects' data differences within the same condition e.g. PRE/POST data differences, and therefore repeated measures. 

• BETWEEN SUBJECT DESIGN is because the study focuses on the subjects' data differences in DIFFERENT CONDITIONS e.g. CONTROL GROUP versus EXPERIMENTAL GROUP, therefore each group is independent of each other.  

• INDEPENDENT SAMPLE has no relation to each other e.g. one person eats fruit  while the other does not. 

• MATCH PAIR SAMPLE has a possible relation to each other e.g. all people eat fruit then all fast.

• The main INDEPENDENT VARIABLE [IV] investigated can have 1+ levels depending on the complexity of the experiment. 

• Simple experiments will have only one [IV] but more complex experiments will have 1+ [IV]. 

• The number of LEVELS within the [IV] will impact the types of STAT TEST you can use on any data obtained from an experiment. 

• So what makes EXPERIMENTAL DESIGNS so scientific, valid, reliable and measurable to REAL LIFE QUESTIONS????? 

• You should understand some KEY STAT terms: "a population" can be just your small sample or alternatively referring to the WIDER WORLD

• The AIM of ANY STUDY is to CHARACTERISE a study's DATA. 

• PUT SIMPLY, how to infer from an experimental data to the wider world... about ANYTHING we are interested in! 

• POPULATION PARAMETERS can be an expressed STAT in 1+ way as per below.

• ALL suggests the level of significance of the data and therefore the CHARACTERISTICS of the DATA in a numerical manner. 

• IF SIGNIFICANT this means we can have a high level of probability that a study's findings - whatever it may be - can be applied to the wider world! 

• LOTS of SYMBOLS are applied in STATS to represent the CHARACTERISTICS within the complex statistical formulas SPSS uses, as per below. 

BUT WHY SHOULD WE BOTHER knowing this at all? 

Can you think of any REAL LIFE IMPLICATIONS? 

• There are TWO MAIN CATEGORIES for STAT TESTS: they can fall into either PARAMETRIC or NON-PARAMETRIC. 

• NON-PARAMETRIC tests are the alternative when PARAMETRIC TESTS are not suitable. 

• AN EASY BREAKDOWN to figure out which one is most appropriate depends on the experimental design & suitable assumptions about the data.  

• There are PRO'S & CON'S of either approach, but the most salient aspect is what is the most suitable method to use for your experiment.

• 1+ STATISTIAL TESTS one can use, however, this blog post will focus just an overview of the T-TEST. 

• Later blog posts will detail the summary table below. 

T-TESTs can FALL in to 2 MAIN TYPES: 

BUT WHY SHOULD WE USE A T-TEST when trying to infer probable conclusions from a population? 

WHAT ABOUT the Z-TEST? Can you think of a condition when it would be more suitable? 

• Z-TESTs are only used when the population parameters are KNOWN within the experiment. 

• T-TESTs are ONLY USED when an experimenter DOES NOT know the POPULATION PARAMETERS!

• T-TESTs can be classed as INDEPENDENT (refer to previous slides if you have forgotten the multi+ names for these two main EXPERIMENTAL DESIGNS). 

• When you finally get around to actually using STAT PACKAGES e.g. SPSS it will look like the box below: 

• The 'ANALYSE' tab on SPSS shows all of the possible STAT TESTS  a person can do to the data obtained. 

• T-TESTs can be classed as DEPENDENT (refer to previous slides if you have forgotten the multi+ names for these two main EXPERIMENTAL DESIGNS). 

• SIGNAL/NOISE within a study refers to things that make it hard to see what our data is really telling us. 

• IN/DEPENDENT T-TESTs will have TWO STANDARD DISTRIBUTIONS = DATA SETS. 

• T-TESTs can be classed as DEPENDENT, in which both data sets are linked and dependent on each other. 

• For example, this could include measuring everyone's IQ after exposing everyone to a new super pill. The difference BEFORE/AFTER will be the two sets of data! 

• There are a few stages that SCIENTISTS SHOULD follow when establishing an EXPERIMENTAL DESIGN towards selecting a STAT TEST.  

• DONT WORRY if you don't understand all of the stages below as this post focuses on introducing the T-TEST & EXP DESIGN. 

• T-TESTs are a very basic measurement that is very useful to compare ANY TWO SETS of DATA.... many experiments can use them to answer real life questions. 

• T-TESTs can only be conducted on DATA if it follows a certain criteria and, therefore, certain assumptions.

• PUT SIMPLY, it can be broken down into 3 STAGES or RULES, which are as follows: 

• T-TESTs MUST SHOW HOMOGENEITY of VARIANCE, aka they must have similar standard distributions.

• There are MANY ALT TESTS people can use - not just a T-TEST.  But this will be blogged about at a later date... baby steps people. 

• The whole rationale for DIFFERENT STAT tests is kinda obvious..... 

• There are 1+ EXP DESIGNS, HYPOTHESES, METHODS, SAMPLES and WAYS in general to measure questions in the world. 

• T-TEST are just one of the simplest measures, whilst ANOVA'S are more complex. 

• For instance, you could use a "Regressional Analysis" which allows you to create a MATH MODEL for predictions based on correlational alone. This is goos for marking masses of information and seeing if the data demonstrates any possible links. 

OVERALL, by learning the RULES of STAT TESTs you can independently value a study's findings for yourself. 

It may seem long winded, but in the end it does benefit you a lot to think more critically. 

STATISTICAL ANLYSIS MADE SIMPLE (part 1.0)

SPSS or ALT STATs PROGRAMME MADE SIMPLE: for PSYCHOLOGY STUDENTS 

Know little to nothing about STATS/MATHS? e.g. avoided A-LEVEL [MATHS] but wanted to study science, ahem? 

HOPEFULLY this will break down [why] & [how] PSYCHOLOGISTS or SOCIAL SCIENTISTS must use statistical mathematics to prove theories (to a probable degree as nothing is 100% proven).

BEFORE you can understand S.P.S.S (a statistical programme that informs psychologists about the usefulness of their results) ..... YOU MUST understand a bit of basic mathematical concepts e.g. MATHS MODELs used in SOCIAL SCIENCE.  

POINT 1: 

Scientists form a MATH MODEL from all of the data that they have collected. 

POINT 2: 

This MATH MODEL is like a theory or hypothesis that can be proven or disproved. 

POINT 3: 

The need for a MATH MODEL, alongside any theory/hypothesis, is because this CAN BE objectively measured & is an UNIVERSAL measurement tool.

HYPOTHESIS: 

If an average joe started to use LYNX deodorant would angels resembling a victoria's secret model start CRASHING DOWN?

 
EXPERIMENT DATA:

Undergoes statistical analysis, e.g. SPSS, which finds that this is an unlikely probability in the real world. 

THEORY: 

The hypothesis of using LYNX is proven most unlikely true, therefore no theory can be conclusive to guarantee hotties being produced on the application of LYNX. 

ANY STATs PACKAGE that has been proven by mathematical geeks to prove a high level of accuracy in probability has many key words to explain why it actually works e.g. MEAN, VARIANCE & STANDARD DEVIATION etc... These are words for mathematical formulas that help scientists to predict future truths to a probable degree. 

So what are the STAGES that scientists undertake from doing the experiment to claiming that the findings appear to be facts .....

e.g. SMOKING causes CANCER [VS] SMOKING does NOT cause CANCER?

FREQUENCY DISTRIBUTIONS are just graphical mathematical representations of all or relevant data collected. This allows the bridge to be crossed between transferring reported things in the world into a mathematical format..... WHY???? It allows us to test whether what a scientist has collected is actually true. 

I can measure happiness levels by asking people, then allocate numbers to each level of happiness, allowing the bridge between verbal happiness to equal numerical values 

e.g. VERY HAPPY = 10 or VERY UNHAPPY = 1. 

Once a scientist can see all of the information they have collected from their study, a graphical representation holistically can tell us a lot e.g. DATA means NOTHING or THERE IS SOME sort of PATTERN. 

FINALLY, a STAT CALCULATION package is used, such as SPSS (there are other types). This will indicate whether what was seen in the graph has any real use e.g. finding that a correlation in the graphical data may be very weak when you test it therefore suggests that it's not very probable to have any predictive powers e.g. LYNX usage does not have any significant predictive effects for the production of a hottie. 

The [RANGE] is a simple measurement of the DATA’S VARIABILITY by just calc the difference between HIGHEST & LOWEST numbers. 

The [RANGE] is limited in its scientific predictive abilities as it ONLY USES 2 NUMBERS & not all the data! That is why the measure of VARIANCE is IMPORTANT. 

[VARIANCE] is showing the SPREAD of ALL the data points COLLECTIVELY. 

MEAN: a simple average value of all data points. 

RANGE: a simple spread value of all data points. 

[STANDARD DEVIATION] is a detailed dispersion value BETWEEN data points.  

BEFORE you can FULLY understand the complex SCIENCE terms [OR] all the different STATs TEST psychologists & social scientists use, you first need to understand some basics concepts to explain how it relates to establishing predictive effects about the REAL WORLD. 


PUT SIMPLY: how does a EXPERIMENT measuring human behaviour then emerge a predictive effect e.g. a scientist states that people who like the colour red tend to be more lazy from various data, from brain waves to behaviour observed (FYI: this is a totally made up experimental finding... but you get the point). 

BASIC STAT KNOWLEDGE is a key foundation, which I am sure many understand, but can you really understand how a bunch of graphical data on a diagram helps scientists explain how spraying LYNX can be argued to find random angels falling from the sky in one study (financed by Lynx) but not in another (financed by the government)? 

These FREQUENCY DISTRIBUTIONS are an ESSENTIAL starting point from any VARIABLE we want to measure and visually see in life towards a MATHEMATICAL MODEL. 

DATA when placed into a GRAPH is a PICTORIAL REPRESENTATION of how VARIOUS MATH FORMULAS are used along the DATA to find its predictive effects. 

BY LOOKING at the VISUAL REPRESENTATION alongside the MATH FORULA OUTPUT, if both MATCH to say a SIGNIFICANT or NON-SIGNIFICANT finding is present, it can be more easily argued that a scientist's data is indeed VERY PROBABLE e.g. LYNX means ANGELS WILL FALL under the right conditions (such as a commercial production!)

WHY CARE about all the DIFFERENT FORMS of graphical representation? Well, as science grows by asking more complex questions about the world, DIFFERENT GRAPHS & STAT TEST may also grow to match the suitability of what TESTS to use on specific DATA. 

PUT SIMPLY: you would not put diesel in a car that should use petrol as the effects would be not valid! 

VARIANCE is also known as the MAIN WORD that measures the LEVEL of DISPERSION. Put simply, how far/close is the data to the MEAN. The basic way to measure the MEAN is learnt at secondary school, but at a higher STATs level, it can also be measured relatively in 1+ ways. 

YOU DO NOT NEED to know EVERY FORMULA off by heart! That's why PSYCHOLOGISTS USE MATH STAT PACKAGES. 

 Psychologists are not MATHEMATICIANS; they understand the concepts, but they do not spend 24/7 on formulas etc.... They spend their time finding answers to the world measuring all the things in the world that we can measure.... They tend to use 1+TOOLS,  e.g. STAT TEST, to help humans shape & understand the world, which the naked eye may fail to see by using MATHS as the universal language to translate it into something we can understand. 

DO NOT get HUNG UP on the [Z]-VALUE in isolation too much. It's good to understand what it can be used for and WHY; but the IMPORTANT THING to bear in mind is that you should consider all of the possible STAT VAULES that can emerge from the STAT TESTS collectively. 

I THINK is is essential to UNDERSTAND when you LOOK AT ANY DATA, the numbers the STATS PACKAGES punches out can be understood in basic terms for ALL VALUES. 

I know it is tempting for undergraduates to just look at the P-VALUE (explained later), alongside the LEVEL of SIGNIFICANCE..... but there are many other forms of data that can also give you a clue to the bigger picture of what the DATA may actually be trying to scream out (links to TYPE I & II ERRORS explained later in the blog). 

HOPEFULLY, by looking at HOW the [Z] formula forms the [Z] value in the STAT TEST you can understand WHY it is important & why IT IS NOT EXACTLY the same as another VARIANCE MEASURE. 

BUT WHAT IS THE POINT of measuring a FORM of the STANDARD DEVIATION [S.D] known as the [Z]-VALUE? 

So hopefully you will understand when someone in class states a variable is above or below by a certain amount e.g. +/- 7 [S.D] FROM THE MEAN....  you will see the relevance and impact on a study's results. 

THERE ARE MANY OTHER VALUES, TERMS & TESTS within STATs; however, please bear in mind that you need the foundations or basic understanding. 

YOU COULD JUST click away AT STAT TESTS e.g. SPSS without really understanding what you are doing and following some book demonstrating how to do a T-TEST. 

The issue with this is that when you have to actually explain your answer in a SCIENTIFIC PAPER you need to understand what all the fun formulas and numbers are really telling you. 

PUT SIMPLY: you should have now a basic understanding of what STAT TESTS are & WHY they are needed in ANY SCIENTIFIC RESEARCH....

... there's no point measuring a bunch of numbers in a person's brain waves so to speak if you can't relate it to real life behaviour or anything relevant to real life.... hence STAT TESTS tries to explain the many wonders of the world.... 


(NB: further notes on SPSS & STATS will be blogged. 
Please bear with me and thanks for our patience)