Primary

 text:  
§  King,  B.M.  &  Minium,  E.W.  (2007).  Statistical  Reasoning  in  the  
Behavioral  Sciences  (5th  Ed.).  Noida:  John  Wiley.  
 
Additional  References:  
§  Garrett,  H.E.  &  Woodworth,  R.S.  (1987).  Statistics  in  Psychology  and  
Education.  Mumbai:  Vakils,  Feffer  &  Simons  Pvt.  Ltd.    
§  Mangal,  S.K.  (2012).  Statistics  in  Psychology  and  Education  (2nd  Ed.).  
New  Delhi:  Prentice  Hall  of  India  
§  What  is  Psychological  Research?  Relevance  of  
Statistics  in  Psychological  Research,  Descriptive  and  
Inferential  Statistics  
§  Scales  of  Measurement  (Nominal,  Ordinal,  Interval,  
Ratio)  
§  Graphical  Representation  of  Data  (Histogram,  
Frequency  Polygon,  Cumulative  Percentage  Curve,  
Bar  Diagram,  Pie  Chart)  
What  is  Psychological  
Research?  
§  Defined  as  “the  scientific  study  of  behaviour,  
mental  processes  and  experiences  of  
individuals”.  
§  About  answering  questions.  
§  Whose?  Yours!  
§  Does  stress  influence  performance?  
§  Does  online  communication  affect  well-­‐being?  
§  Does  exposure  to  media  images  influence  body  image?  
 
A  systematic,  empirical,  critical  investigation  that  is  
structured  to  answer  questions  about  the  behaviour  and  
experiences  of  individuals.  

Goals  of  psychological  research:  

§  Description  
§  Explanation  
§  Prediction  
§  Control    
§  Application  
 
Clinical Research  on  causes  of  disorders  and  
treatment  effectiveness
Developmental Study  of  physical,  emotional,  cognitive  
and  social  development  across  the  entire  
life  span
Cognitive Study  of  mental  processes  such  as  
memory,  problem  solving,  reasoning  etc.
Experimental   Research  on  basic  processes  such  as  
learning,  perception  &  motivation
Personality   Study  of  individual  differences  in  personality  
  and  their  effects  on  behaviour  
Social   Examination  of  how  the  social  environment  
influences  an  individual’s  behaviour  
Physiological   Study  of  relationship  between  biological  
processes  and  behaviour  
Cultural/   Study  of  cultural  transmission,  psychological  
Cross-­‐cultural   similarities  &  differences  among  people  from  
different  cultures  
Organizational   Examination  of  behaviour  in  work  settings  
Sports   Analysis  of  the  performance  of  athletes  &  
the  use  of  psychological    principles  to  help  
them  achieve  peak  performance  levels  
Forensic   Application  of  psychological  knowledge  to  
human  problems  in  the  field  of  law  
enforcement  
Health   Research  on  psychological  factors  that  
influence  the  origin,  prevention  and  
treatment  of  physical  illness  
§  Basic  Research:  Aims  to  examine  how  and  why  
people  behave,  think  &  feel  they  way  they  do.    

§  Applied  Research:  Aims  to  use  basic  scientific  

knowledge  to  solve  specific,  practical  problems.    
§  Exploratory  Research:  Research  whose  primary  purpose  is  to  
examine  a  little  understood  issue  or  phenomenon  and  to  
develop  preliminary  ideas  about  it  and  move  toward  refined  
research  questions.    
§  Descriptive  Research:  Research  in  which  the  primary  purpose  is  
to  “paint  a  picture”  using  words  or  numbers  and  to  present  a  
profile,  a  classification  of  types,  or  an  outline  of  steps  to  answer  
questions  such  as  who,  when,  where,  and  how.  
§  Explanatory  Research:  The  primary  purpose  is  to  explain  why  
events  occur  and  to  build,  elaborate,  extend,  or  test  theory.  
Basis  of   Quantitative  Research   Qualitative  Research  
Difference      
Approach   Emphasis  is  on  a  positivist  approach,   Emphasis  is  on  an  interpretivist  approach  
stressing  on  hard  data  often  in  the   stressing  on  soft  data  (words,  impressions,  
form  of  numbers.     symbols).      

Research   Clearly  set  linear  path  and  successive   Non  linear  research  path  that  permits  and  
Path   procedures  that  seem  to  follow  in  a   obligates  the  researcher  to  go  in  cyclical,  
logical  sequence.     back  and  forth,  and  non-­‐successive  
sequences.  

Sample   Usually  a  small  group  of  respondents.   Usually  a  large  number  of  cases  
representing  a  population  of  interest.  

Research   Questions  are  finalized  before  the   May  start  out  with  a  vague  or  poorly  
Question     study  and  are  used  in  developing   defined  research  question  which  may  
evolve  as  the  study  progresses  and  new  
steps  and  guiding  the  study.  
insights  are  gained  and  incorporated.
     
Basis  of   Quantitative  Research   Qualitative  Research  
Difference    
Research   Use  highly  structured  methods   Use  semi-­‐structured  methods  such  as  in-­‐
methods   such  as  questionnaires,  surveys   depth  interviews,  focus  groups  and  
and  structured  observation.   participant  observation.  

Purpose  of   Primarily  deductive;  used  to  test   Primarily  inductive;  used  to  formulate  theory  
research   pre-­‐specified  concepts,   or  hypotheses.  
constructs,  and  hypotheses  that  
make  up  a  theory.  

Focus     Stress  is  on  objectivity.   Allows  subjectivity;  researchers  can  employ  
personal  insights,  feelings,  and  human  
perspectives  to  understand  social  life.    

Analysis   Analysis  proceeds  by  using   Analysis  proceeds  by  summarizing,  

statistics,  tables,  or  charts  and   categorizing  and  interpreting  data  to  provide  
discussing  how  the  results  relate   valuable  insights  that  are  open  to  exploration.
to  hypotheses.          
§  Initial  observation  or  research  question    
§  Gather  information  &  form  hypothesis    
§  Select  the  research  method  &  design  the  study  
§  Collect  data    
§  Analyze  the  data  and  draw  conclusions  
§  Report  findings  to  scientific  community  
§  Further  research  &  theory  building  
§  New  hypotheses  derived  from  theory  
Relevance  of  Statistics  In  
Psychological  Research  
§  Statistics  is  a  branch  of  applied  mathematics  that  
focuses  on  the  organization,  classification,  analysis  and  
interpretation  of  data.    
§  Statistics  are  all  around  you  –  from  your  college  grade  
point  average  (GPA)  to  a  poll  predicting  which  political  
candidate  is  likely  to  win  an  election.  
§  In  each  case,  statistics  are  used  to  inform  you.  The  
challenge  is  to  be  able  to  identify  statistics  and  to  
interpret  what  they  mean.    
Psychologists  use  statistical  techniques  for  the  following  
functions:  
§  Organize  data:  When  dealing  with  an  enormous  amount  of  
information,  it  is  all  too  easy  to  become  overwhelmed.  
Statistics  allow  psychologists  to  present  data  in  ways  that  are  
easier  to  comprehend.    
§  Visual  displays  such  as  graphs,  pie  charts,  frequency  
distributions  and  scatter  plots  make  it  possible  for  researchers  
to  get  a  better  overview  of  the  data  and  to  look  for  patterns  
they  might  otherwise  miss.    
 
§  Summarize  data:  Statistics  is  partly  about  describing  
observations  as  accurately  and  succinctly  as  possible.    
§  This  is  achieved  by  using  formulae  which  turn  fairly  
complex  information  into  simple  indices  that  describe  
numerically  its  main  features.    
§  These  indices  include  measures  of  central  tendency  and  
variability  that  help  in  summarizing  and  comprehending  a  
large  set  of  data.  

 
§  Determine  the  extent  of  relationships  among  variables:  
Most  psychological  research  is  intended  to  examine  
relationships  between  two  or  more  variables.    
§  Using  statistics,  researchers  can  measure  the  direction  and  
strength  of  the  relationship  between  variables  as  they  exist  
naturally.    
§  Make  inferences  based  upon  data:  By  using  what’s  known  as  
inferential  statistics,  psychologists  can  analyze  and  make  
inferences  from  the  data  to  reach  valid  conclusions  that  extend  
beyond  the  immediate  data.    
§  This  is  important  because  we  observe  only  a  sample  of  elements,  
and  reason  from  the  particular  facts  or  cases  to  draw  general  
conclusions  –  a  process  called  as  induction.    
§  Thus,  statistics  help  the  researcher  to  answer  the  questions  that  
initiated  the  research  by  determining  exactly  what  general  
conclusions  are  justified  based  on  the  specific  results  that  were  
obtained.    
§  Statistics  consists  of  univariate  and  multivariate  procedures.    
§  Psychologists  use  univariate  procedures  when  they  measure  
only  one  variable.  
§  They  use  multivariate  procedures  when  multiple  variables  
are  used  to  ascertain  the  relationship  between  two  or  more  
variables,  to  make  inferences  from  the  data,  to  extract  
factors  (or  latent  variables)  etc.      
§  A  personality  psychologist  may  use  statistics  to  study    
individual  differences  in  personality  and  their  effects  on  
behaviour.  
§  A  family  counselor  may  use  statistics  to  describe  patient  
behavior  and  the  effectiveness  of  a  treatment  program.  
§  A  sports  psychologist  may  use  statistics  to  analyze  the  
performance  of  athletes.    
§  A  cross-­‐cultural  psychologist  may  use  statistics  to  study  
differences  among  people  from  different  cultures.    
§  The  reason  it  is  important  to  study  statistics  can  be  
described  by  the  words  of  Benjamin  Disraeli,  19th-­‐century  
British  statesman:  “There  are  lies,  damned  lies  and  
statistics.”    
§  He  meant  that  statistics  can  be  deceiving  –  and  so  can  
interpreting  them.    
§  The  purpose  of  descriptive  statistics  is  to  organize  and  to  
summarize  observations  so  that  they  are  easier  to  
comprehend.    
§  Descriptive  statistics  is  important  because  if  we  simply  
presented  our  raw  data  it  would  be  hard  to  visualize  or  
understand  what  the  data  was  showing,  especially  if  there  
was  a  lot  of  it.    
§  Descriptive  statistics  therefore  enables  us  to  present  the  
data  in  a  more  meaningful  way,  which  allows  simpler  
interpretation  of  the  data.    
§  Descriptive  statistics  makes  use  of  tabular,  graphical  and  
numerical  techniques  to  present  the  data.  
§  It  is  used  to  describe  data  in  terms  of:  
§  Distribution  (frequency  tables,  bar  graphs,  pie  charts  
etc.)  
§  Center  (measures  of  central  tendency:  mean,  median  
mode)  
§  Spread  (measures  of  variability:  range,  standard  
deviation,  variance)  
   
Some  examples:  
§  A  teacher  summarizes  the  general  performance  of  a  student  
across  a  wide  range  of  course  experiences  using  the  Grade  Point  
Average  (GPA).  
§  A  sports  fan  ranks  a  team’s  players  according  to  their  batting  
averages.    
§  A  college  professor  administers  a  test  of  statistical  competence  
and  summarizes  the  scores  to  interpret  the  competence  level  of  
her  students.    
§  An  organization  obtains  the  average  number  of  sales  made  by  
its  telemarketing  executives  to  assess  their  performance.    
 
§  The  purpose  of  inferential  statistics  is  to  draw  a  conclusion  
(an  inference)  about  conditions  that  exist  in  a  population  (the  
complete  set  of  observations)  from  the  study  of  a  sample  (a  
subset)  drawn  from  the  population.    
§  Population  includes  all  observations  (e.g.  students’  scores,  
people’s  incomes,  etc.)  in  which  the  researcher  is  interested.    
§  Sample  is  a  carefully  chosen  subset  of  the  population.  We    use  
the  sample  to  infer  something  about  the  characteristics  of  
the  population.    
§  Typically,  the  research  process  begins  with  a  question  about  
a  population  parameter.  However,  data  is  obtained  from  a  
subset  of  the  population  and  a  sample  statistic  is  computed  
from  the  data  to  make  an  inference  about  the  population  
parameter.  
§  Parameters  are  the  real  entities  of  interest,  and  the  
corresponding  statistics  are  guesses  at  reality.    
§  Thus,  we  want  to  infer  something  about  the  characteristics  
of  the  population  (parameter)  from  what  we  know  about  the  
characteristics  of  the  sample  (statistic).    
 
Fig: The steps used in statistical inference
Some  examples:    
§  A  pollster  asks  a  group  of  voters  how  they  intend  to  vote  in  the  
upcoming  elections.    
§  A    medical  researcher  examines  the  claim  that  large  doses  of  
vitamin  C  can  help  prevent  the  common  cold.    
§  A    social  psychologist  evaluates  the  influence  of  media  exposure  
on  body  image  among  teenage  students.    
§  A  researcher  tests  a  new  diet  drug  on  a  group  of  overweight  
individuals.    
§  A  personality  psychologist  studies  gender  differences  on  
emo=onal  expressiveness.    
Descriptive  Statistics   Inferential  Statistics  
Aim  of  descriptive  statistics  is  to   Aim  of  inferential  statistics  to  draw  
summarize  the  current  dataset.   conclusions  about  a  population  
outside  of  the  obtained  dataset.  
Descriptive  statistics  usually   Inferential  statistics  takes  a  sample  
operate  within  a  specific  area  that   of  a  population  especially  if  the  
contains  the  entire    target   population  is  too  big  to  conduct  
population.   research,  or  when  we  don’t  have  
    access  to  the  entire  population.    
It  does  not  allow  us  to  make   It  allows  us  to  make  conclusions  
conclusions  beyond  the  data  we   beyond  the  immediate  data  we  have  
have  analyzed.     analyzed.    
Please note: Even when a data analysis draws its main conclusions using
inferential statistics, descriptive statistics are also generally presented.
SCALES  OF  
MEASUREMENT  
§  Measurement  is  a  process  of  systematically  assigning  values  
(numbers  or  scores)  to  properties  of  people,  places,  things,  or  
events.    
§  The  way  the  values  are  assigned  determines  the  level/scale  of  
measurement.  It  refers  to  the  amount  of  information  the  
measurement  procedure  can  convey  about  the  actual  quantity  
of  the  variable  present  and  about  the  differences  in  individuals  
with  different  scores.    
§  The  levels  of  measurement  differ  both  in  terms  of  the  meaning  
of  the  numbers  used  in  the  measurement  system  and  in  the  
types  of  statistical  procedures  that  can  be  applied  appropriately  
to  data  measured  at  each  level.  
§  Nominal  scale  is  the  simplest  of  the  four  levels  of  measurement.      
§  It  involves  the  process  of  placing  observations  into  categories  
that  differ  in  some  qualitative  aspect.    
§  The  categories  used  for  classification  must  be  mutually  
exclusive  (observations  cannot  fall  into  more  than  one  category)  
and  exhaustive  (there  must  be  enough  categories  for  all  the  
observations).    
§  Variables  that  are  qualitative,  or  categorical  in  nature  are  
usually  measured  on  a  nominal  scale,  because  we  merely  assign  
category  labels.  
§  For  example,  gender,  political  affiliation,  or  eye  color  are  
qualitative  variables  that  can  be  measured  on  a  nominal  scale.    
§  With  a  nominal  scale,  there  is  no  question  about  one  
category  having  more  or  less  of  any  particular  quality;  all  
categories  are  simply  different.  
§  Although  the  categories  on  a  nominal  scale  are  not  
quantitative  values,  they  are  occasionally  represented  by  
numbers.  However,  the  numbers  are  only  arbitrary  and  do  
not  designate  “more”  or  “less”  of  anything  (e.g.,  1  =  Red,  2  =  
Blue,  3  =  Green,  4  =  Yellow).  
§  For  example,  the  rooms  or  offices  in  a  building  may  be  
identified  by  numbers.  But  the  room  numbers  are  simply  
names  and  do  not  reflect  any  quantitative  information.  Room  
109  is  not  necessarily  bigger  than  Room  100  and  certainly  not  
9  points  bigger.    

§  It  also  is  fairly  common  to  use  numerical  values  as  a  code  for  
nominal  categories  when  data  are  entered  into  computer  
programs.    
§  For  example,  the  data  from  a  survey  may  code  males  with  a  0  
and  females  with  a  1.  Again,  the  numerical  values  are  simply  
names  and  do  not  represent  any  quantitative  difference.    
§  At  the  next  level  of  complexity  is  the  ordinal  scale  (the  Latin  
root  means  “order”).  In  this  type  of  measurement,  the  
categories  must  still  be  mutually  exclusive  and  exhaustive,  but  
they  also  indicate  the  order  of  magnitude  of  some  variable.  
§  With  a  nominal  scale,  the  outcome  of  classification  is  a  set  of  
unordered  categories.  With  the  ordinal  scale,  it  is  a  set  of  
ordered  categories  or  ranks.  
§  Thus,  ordinal  scale  constitutes  a  form  of  ordering  or  ranking  of  
responses  along  some  underlying  dimension  that  expresses  
“more”  or  “less”  of  something.    
§  An  example  is:  instructor,  assistant  professor,  associate  
professor,  and  professor.  

§  In  another  example,  a  supervisor  may  estimate  the  

competence  of  seven  workers  by  arranging  them  in  order  of  
merit.  The  only  relation  expressed  in  a  series  of  numbers  used  
in  this  way  is  that  of  “greater  than.”    

§  We  may  use  numbers  for  the  ranks  but  it  is  not  necessary.  
Among  persons  ranked  1,  2,  and  3,  the  first  person  has  a  
greater  degree  of  merit  than  the  person  ranked  second,  and  
the  second  person  has  greater  merit  than  the  third.  
§  However,  the  interval  between  two  successive  ranks  is  
indeterminate.  Consequently,  the  difference  between  any  
two  consecutive  ranks  (e.g.  ranks  1  and  2)  may  not  be  the  
same  as  that  between  another  pair  of  consecutive  ranks  (e.g.  
ranks  2  and  3).  Thus,  ordinal  measurements  describe  order  
but  not  the  relative  size  or  degree  of  difference  between  the  
adjacent  steps  on  the  scale.  

§  Also,  nothing  is  implied  about  the  absolute  level  of  merit.  All  
observations  may  denote  excellence,  or  they  could  be  just  be  
average.    
§  The  next  major  level  of  complexity  is  the  interval  scale.  
§  This  scale  has  all  the  properties  of  the  ordinal  scale,  but  with  
the  further  refinement  that  a  given  interval  (distance)  
between  scores  has  the  same  meaning  anywhere  on  the  
scale.    
§  Thus,  interval  scale  not  only  tells  us  about  the  ordering  of  
observations  but  also  indicates  distance  between  them.    
§  It  allows  us  to  know  how  many  units  greater  than,  or  less  
than,  one  observation  is  from  another  on  the  measured  
characteristic.    
§  Examples  of  this  type  of  scale  are  degrees  of  temperature  on  
the  Fahrenheit  or  Celsius  scales.  A  10° rise  in  a  reading  on  the  
Celsius  scale  represents  the  same  change  in  heat  when  going  
from  0°  to  10°  as  when  going  from  20°  to  30°.
§  However,  zero  point  on  interval  scales  is  an  arbitrary  reference  
point;  the  value  of 0 is  assigned  to  a  particular  location  on  the  
scale  simply  as  a  matter  of  convenience  or  reference.  Thus,  a  
value  of  zero  does  not  indicate  a  total  absence  of  the  quality  
being  measured.  In  the  Celsius  scale,  0°C  is  the  freezing  point  
of  water  and  does  not  imply  an  absence  of  heat.  
§  Also,  it  is  not  possible  to  speak  meaningfully  about  a  ratio  
between  two  measurements  on  an  interval  scale.  Therefore,  
we  cannot  assert  that  a  temperature  of  100°  Celsius  is  twice  as  
hot  as  one  of  50° ;  or  that  a  rise  from  90° to  99° Celsius  is  a  
10% increase.    
§  A  ratio  scale  possesses  all  the  properties  of  an  interval  scale  
and  in  addition  has  an  absolute  zero  point  in  which  there  is  
total  absence  of  the  characteristic  being  measured.  
§  It  is  therefore  possible  to  speak  meaningfully  about  a  ratio  
between  two  measurements.    
§  The  Kelvin  scale  has  an  absolute  zero,  the  point  at  which  a  
substance  would  have  no  molecular  motion  and,  therefore,  
no  heat.  Thus,  100°  is  twice  as  hot  as  50° on  the  Kelvin  
scale.  
§  Other  examples  of  ratio  scale  measurements  are  length,  
weight,  and  measures  of  elapsed  time.  
§  Not  only  is  the  difference  between  40  in.  and  41  in.  the  same  
as  the  difference  between  80  in.  and  81  in.,  but  it  is  also  true  
that  80  in.  is  twice  as  long  as  40  in.  

A Ratio Scale
Each  scale  of  measurement  sa=sfies  one  or  more  of  the  following  
proper=es  of  measurement.  
§  Identity.  Each  value  on  the  measurement  scale  has  a  unique  
meaning.  
§  Magnitude.  Values  on  the  measurement  scale  have  an  ordered  
rela=onship  to  one  another.  That  is,  some  values  are  larger  and  
some  are  smaller.  
§  Equal  intervals.  Scale  units  along  the  scale  are  equal  to  one  
another.    
§  Absolute  zero.  The  scale  has  a  true  zero  point,  below  which  no  
values  exist.  
§  In  a  nominal  scale,  numbers  are  assigned  to  categories  as  
"names".  Which  number  is  assigned  to  which  category  is  
completely  arbitrary.  Therefore,  the  number  only  gives  us  the  
identity  of  the  category  assigned.  
§  Ordinal  scales  have  the  property  of  magnitude  as  well  as  
identity.  The  numbers  represent  a  quality  being  measured  
(identity)  and  can  tell  us  whether  a  case  has  more  of  the  
quality  measured  or  less  of  the  quality  measured  than  another  
case  (magnitude).  The  distance  between  scale  points  is  not  
equal.  
§  The  interval  scale  of  measurement  has  the  properties  of  
identity,  magnitude,  and  equal  intervals.  With  an  interval  
scale,  you  know  not  only  whether  different  values  are  bigger  or  
smaller,  you  also  know  how  much  bigger  or  smaller  they  are.    
§  The  ratio  scale  of  measurement  satisfies  all  four  of  the  
properties  of  measurement:  identity,  magnitude,  equal  
intervals,  and  an  absolute  zero.  
§  In  the  behavioral  sciences,  there  are  many  measuring  instruments  
that  lack  equal  intervals  and  an  absolute  zero  point.    
§  Consider,  for  example,  a  spelling  test.  Items  might  be  words  such  
as  garden,  baseball,  and  rowboat.  A  score  of  zero  on  this  test  
means  that  the  person  could  not  spell  the  simplest  word  on  the  
list,  but  what  if  simpler  words  had  been  on  the  test,  such  as  cat,  
run,  and  bat?    
§  Our  spelling  test,  then,  does  not  have  an  absolute  zero  point  
because  zero  on  the  spelling  test  does  not  indicate  a  total  absence  
of  spelling  ability.    
§  The  same  is  true  of  midterm  tests,  IQ  tests,  the  SAT,  and  almost  all  
other  tests  of  mental  performance.  
§  What  about  equal  intervals?  To  have  equal  intervals  on  our  spelling  
test,  we  should  be  able  to  state  quantitatively  just  how  much  more  
spelling  ability  is  needed  to  spell  garden  than  to  spell  cat.  
§  However,  there  is  no  objective  way  to  determine  when  equal  
numerical  intervals  on  a  mental  test  represent  equal  increments  in  
performance.  
§  Hence,  some  people  argued  that  calculating  certain  statistical  
variables  (such  as  averages)  on  tests  of  mental  abilities  could  be  
seriously  misleading.  
§  Fortunately,  the  weight  of  the  evidence  suggests  that  in  most  
situations,  making  statistical  conclusions  is  not  seriously  hampered  
by  uncertainty  about  the  scale  of  measurement.  
§  However,  there  are  several  areas  where  we  need  to  be  aware  of  
scale  problems  to  avoid  taking  tempting  but  erroneous  
positions.  
§  For  example,  we  should  not  say  that  a  person  with  an  IQ  of  150  
is  twice  as  bright  as  one  with  an  IQ  of  75.  
§  Similarly  we  should  not  assume  that  the  difference  between  15  
and  25  points  on  a  spelling  test  necessarily  represents  the  same  
increment  in  spelling  ability  as  the  difference  between  a  score  of
30 and  40  points  on  the  same  test.    
§  In  psychological  measurement,  this  problem  may  be  particularly  
critical  when  a  test  does  not  have  enough  “top”  or  “bottom”  to  
differentiate  adequately  among  the  group  measured.  
§  For  example,  imagine  a  test  of  ability  that  has  a  maximum  possible  
score  of  50  points  and  that  is  too  easy  for  the  group  measured.    
§  For  two  persons  who  score  50  points,  the  score  for  one  may  
indicate  the  maximum  level  of  achievement,  but  the  second  
person  may  be  capable  of  a  much  higher  level  of  performance.    
§  The  measuring  instrument  is  simply  incapable  of  showing  this  
difference  because  it  does  not  include  items  of  greater  difficulty.  
The  level  of  measurement  of  a  variable  tells  us  which  
statistics  are  permissible  and  appropriate.    

  DESCRIPTIVE  STATISTICS    
NOMINAL   ORDINAL   INTERVAL  &  RATIO  
Frequency  tables   Frequency  tables   Frequency  tables  
Mode   Percentiles   Mode    
Mode   Median    
Median   Mean  
Range   Range    
Variance    
Standard  Deviation  
Statistical  tests  are  divided  into  two  types:  parametric  and  
nonparametric  tests.  Parametric  tests  are  more  powerful,  but  
because  they  include  particular  mathematical  operations  on  the  
values,  they  can  be  used  only  with  interval  or  ratio  data.  Ordinal  
and  nominal  data  require  the  use  of  non-­‐parametric  tests.  

INFERENTIAL    STATISTICS  
NOMINAL   ORDINAL   INTERVAL  AND  RATIO  
Non-­‐parametric  tests:   Non-­‐parametric  tests:   Parametric  tests:    
Chi-­‐square  test   Rank-­‐order  correlation   Pearson’s  correlation  
Mann-­‐Whitney  U  test   coefficient,  t-­‐test    
Kruskal-­‐Wallis  test   ANOVA,  Regression,  
Friedman’s  ANOVA   Factor  analysis      
GRAPHICAL  
REPRESENTATION  OF  
DATA  
§  Frequency  distributions  present  the  main  features  of  data  
succinctly,  but  they  are  still  abstract  numerical  
representations  and  require  effort  to  interpret.  
§  Graphs  can  impart  the  same  information  and  speak  to  us  
more  directly  by  pictorially/  visually  presenting  the  
pertinent  features  of  the  data.  
§  Their  ease  of  interpretation  makes  them  particularly  
useful  when  we  want  to  present  data  to  the  general  
public.    
§  A  well  formatted  graph  helps  in  visually  illustrating  certain  
characteristics  and  trends  in  a  set  of  data.  
§  A  study  that  compared  the  use  of  frequency  tables  and  
graphs  in  science  concluded  that  graphs  are  highly  
preferred  because  “of  their  relative  readability,  ease  of  
comprehension,  combinability,  and  overall  rhetorical  
effectiveness”  (Smith  et  al.,  2002).  
§  There  are  many  ways  to  graph  data.  We  will  be  discussing  
the  five  most  common  graphs:  bar  graphs,  pie  charts,  
histograms,  frequency  polygons,  and  cumulative  frequency  
graphs.    
§  Qualitative  variables  are  usually  represented  by  bar  graphs  
and  pie  charts.    
§  Quantitative  variables  are  represented  by  histograms,  
frequency  polygons,  and  cumulative  frequency  graphs.    
§  Graphed  frequency  distributions  generally  have  two  
perpendicular  lines  called  axes:  X-­‐axis  (horizontal  axis,  
abscissa),  Y-­‐axis  (ver=cal  axis,  ordinate).  
§  The  measurement  scale  (set  of  X  values  or  categories)  is  
listed  along  the  X-­‐axis  (with  values  increasing  from  leM  to  
right  for  quan=ta=ve  variables).    
§  The  frequencies  (or  some  func=on  of  frequency)  are  
listed  on  the  Y-­‐axis  with  values  increasing  from  boNom  
to  top.    
§  As  a  general  rule,  the  point  where  the  two  axes  intersect  
should  have  a  value  of  zero  for  both  the  scores  and  the  
frequencies.  If  it  does  not  represent  a  0,  we  must  make  a  
break  in  the  X-­‐axis  to  indicate  that  a  por=on  of  the  scale  is  
missing.  
§  Graph  should  be  constructed  so  that  its  height  (Y-­‐axis)  is  
approximately  three-­‐quarters  (3/4th)    of  its  width  (X-­‐axis).    
§  The  graph  should  have  an  informa=ve  =tle.  
§  Both  the  axes  should  have  appropriate  labels.  
§  A  histogram  is  the  most  commonly  used  graph  to  show  
frequency  distributions.  
§  It  is  a  chart  that  plots  the  frequency  distribution  of  a  
numeric  variable  as  a  series  of  adjacent  bars/rectangles,  
each  of  which  represents  the  scores  in  one  of  the  class  
intervals  of  the  distribution.  
§  The  two  vertical  boundaries  or  the  edges  of  the  bar  
coincide  with  the  real  limits  of  the  particular  class  interval.    
§  The  height  of  a  bar  represents  the  frequency  of  scores  for  
that  class  interval.  We  can  use  either  frequencies  or  
relative  frequencies.  
Steps  in  construc-on  
Step  1:  Construct  a  frequency  distribu=on.  
Step  2:  Decide  on  a  suitable  scale  for  X-­‐axis  by  iden=fying  
and  adding  2  class  intervals  falling  immediately  outside  the  
end  class  intervals.  The  number  of  class  intervals  thus  
obtained  will  be  the  number  of  squares  required  for  the  
width  of  the  graph.  
Step  3:  Decide  on  a  suitable  scale  for  Y-­‐axis  by  mul=plying  
the  width  by  ¾  or  .75  to  find  the  approximate  number  of  
squares  for  the  graph’s  height.    
Step  4:  Draw  bars  of  equal  width  for  each  class  interval  in  such  
a  way  that  the  height  of  a  bar  corresponds  to  the  frequency  or  
rela=ve  frequency  of  that  interval.  There  should  be  no  gaps.  
The  edges  of  bar  represents  both  the  upper  real  limit  of  one  
interval  and  the  lower  real  limit  of  the  next  higher  interval.  
Step  5:  Iden=fy  the  class  intervals  by  using  either  real  limits  or  
mid-­‐points.  If  you  use  real  limits,  place  them  under  the  edge  of  
each  bar.  If  you  use  mid-­‐points,  place  them  under  the  middle  of  
each  bar.  
Step  6:  Label  the  axes  and  give  the  histogram  a  =tle.    
Note:  The  midpoint  can  be  found  by  averaging  the  highest  and  the  lowest  scores  in  the  
interval.    
Results from 50 Students on the History Class Midterm
Examination
Results from 50 Students on the History Class Midterm
Examination
Histogram of grouped history midterm exam scores
§  A  frequency  polygon  is  exactly  like  a  histogram  except  
that  points  are  drawn  rather  than  bars.    
§  These  points  are  plotted  above  the  mid-­‐point  of  each  
class  interval  at  a  height  equal  to  the  frequency  or  
relative  frequency  of  scores  in  that  interval.    
§  The  points  are  then  connected  by  straight  lines.    
§  To  ensure  that  the  graph  is  truly  a  polygon  (i.e.,  the  
graph  is  a  closed  figure),  we  generally  bring  the  straight  
line  down  to  the  mid-­‐points  of  the  additional  first  and  
last  class  intervals  with  zero  frequencies.  
Steps  in  construc-on  
Step  1:  Construct  a  frequency  distribu=on.  
Step  2:  Decide  on  a  suitable  scale  for  X-­‐axis  and  Y-­‐axis.    
Step  3:  Label  the  class  interval  mid-­‐points  along  the  X-­‐axis.  
Step  4:  Place  a  dot  above  the  mid-­‐point  of  each  class  interval  
at  a  height  equal  to  the  frequency  or  rela=ve  frequency  of  
the  scores  in  that  interval.  
Step  5:  Connect  the  dots  with  straight  lines.    
Step  6:  Label  the  axes  and  give  the  polygon  a  =tle.    
Frequency polygon of grouped history midterm exam scores
§  Ordinarily,  we  bring  the  polygon  down  to  the  horizontal  
axis  at  both  ends.    
§  To  do  so,  iden=fy  the  two  class  intervals  falling  
immediately  outside  those  end  class  intervals  containing  
scores.  The  midpoints  of  these  intervals,  ploNed  at  zero  
frequency,  are  then  connected  to  the  graph.  
§  It  was  done  for  the  interval  57–59  because  scores  in  this  
interval  were  possible,  though  not  obtained.    
§  However,  what  do  we  do  when  scores  in  the  next  
adjacent  class  interval  are  not  possible?    
§  The  best  thing  to  do  in  a  case  like  this  is  to  leave  the  dot  
“dangling.”  
§  For  the  interval  96–98  the  graph  is  leM  “dangling”  
because  scores  greater  than  100  were  not  possible  on  the  
history  exam  and  the  next  adjacent  interval  is  99–101.  
§  Bringing  the  polygon  down  to  0  at  the  midpoint  of  the  
interval  99-­‐101  may  mislead  someone  looking  at  the  
graph  to  think  that  scores  in  that  interval  were  possible.  
§  Both  the  histogram  and  the  polygon  are  used  for  
graphing  quan=ta=ve  data  on  an  interval  or  ra=o  scale.  
§  You  can  see  the  similari=es  between  the  two  graphs  by  
superimposing  a  histogram  and  polygon  of  the  same  set  
of  data,  as  shown  in  the  next  slide.  
§  Nevertheless,  there  are  occasions  when  one  may  be  
preferred  over  the  other.  
Frequency polygon of grouped history exam scores
superimposed on a histogram of the same scores
§  A  histogram  is  oMen  used  when  graphing  an  ungrouped  
frequency  distribu=on  of  a  discrete  variable  (or  data  treated  as  a  
discrete  variable).  
§  The  general  public  seems  to  find  a  histogram  a  liNle  easier  to  
understand  than  a  polygon,  and  hence  it  may  be  a  good  choice  
for  communica=ng  with  them.  
§  A  histogram  also  has  some  merit  when  displaying  rela=ve  
frequency.  The  total  area  in  a  histogram  represents  100%  of  the  
scores,  and  thus  the  area  in  the  bars  of  a  histogram  is  directly  
representa=ve  of  rela=ve  frequency.  That  is,  the  area  in  any  
rectangle  is  the  same  frac:on  of  the  total  area  of  the  histogram  
as  the  frequency  of  that  class  interval  is  of  the  total  number  of  
cases  in  the  distribu:on.  
§  However,  representing  frequencies  by  bars  suggests  that  the  
scores  are  evenly  distributed  within  each  class  interval  and  
that  the  borders  of  the  intervals  are  points  of  decided  change.  
The  horizontal  top  of  each  rectangular  bar  is  responsive  only  
to  what  occurs  in  one  class  interval.    Therefore  it  does  not  
represent  a  trend  of  increasing  or  decreasing  frequency  
accurately.  
§  Also,  when  comparing  two  or  more  distributions,  horizontal  
lines  in  the  histogram  will  often  coincide  creating  considerable  
confusion.  Therefore,  it  is  not  used  for  comparisons.  
 
§  A  frequency  polygon  is  often  preferred  for  grouped  
frequency  distribution  of  a  continuous  variable  because  it  
shows  the  gradual  change  over  a  wide  range  of  scores  and  
suggests  continuity  of  the  variable.  
§  The  direction  of  the  straight  line  in  a  frequency  polygon  is  
determined  by  the  frequencies  in  two  adjacent  class  
intervals.  Therefore,  if  a  definite  trend  of  increasing  or  
decreasing  frequencies  exist  over  a  span  of  several  
consecutive  class  intervals,  it  will  represent  this  trend  
directly.    
§  Frequency  polygons  are  particularly  helpful  when  comparing  
two  or  more  distributions.  When  distributions  are  based  on  
different  number  of  cases,  we  can  equalize  that  difference  by  
using  relative  frequencies  rather  than  raw  frequencies.    
 

Relative frequency polygons used to compare two distributions

§  The  Bar  Diagram  is  used  for  depicting  qualitative  
categories  on  a  nominal  or  ordinal  scale  of  
measurement.    
§  It  is  similar  to  histogram  except  that  space  appears  
between  the  rectangles,  thus  suggesting  the  essential  
discontinuity  of  the  several  categories.    
§  However,  within  categories,  subcategories  may  be  
displayed  as  adjacent  bars.    
§  Because  qualitative  categories  on  a  nominal  scale  of  
measurement  have  no  necessary  order,  you  may  
arrange  them  in  any  order  (e.g.,  in  order  of  their  
frequencies,  if  desired).    
§  However,  for  ordinal  scales  of  measurement,  you  
should  arrange  the  categories  in  order  of  rank  (e.g.,  
freshman,  sophomore,  junior,  senior).  
Frequency and relative frequency distribution of
dating apps used by the survey respondents

Dating apps Frequency Relative

frequency (%)
Tinder 153 51.9

Bumble 108 36.6

OkCupid 20 6.8

Hinge 8 2.7
Grindr 6 2.0
Example of a bar diagram for qualitative data with sub-
categories
Frequency distribution of preference for a season among
a sample of 100 students
Seasons f

Winter 72

Autumn 15

Spring 10

Summer 3
n = 100
§  Pie  chart  is  a  circular  graph  whose  pieces  add  up  to  
100%.    
§  Like  the  bar  diagram,  it  is  useful  for  presenting  
qualitative  data.  
§  However,  unlike  the  bar  diagram,  in  which  results  can  
be  expressed  as  either  raw  frequencies  or  relative  
frequencies,  pie  charts  always  use  relative  
frequencies.  
§  The  area  in  any  piece  of  the  pie  is  the  same  fraction  
of  the  pie  as  the  frequency  of  that  category  is  of  the  
total  number  of  cases  in  the  distribution.    
§  In  some  instances,  a  piece  of  the  pie  chart  is  
exploded  (moved  slightly  outward)  to  indicate  that  
section  of  the  pie  is  the  most  noteworthy.    
Frequency and relative frequency distribution of dating apps
used by the survey respondents

Dating apps Frequency Relative

frequency (%)
Tinder 153 51.9
Bumble 108 36.6
OkCupid 20 6.8
Hinge 8 2.7
Grindr 6 2.0
Frequency and relative frequency distribution of preference
for a season among a sample of 100 students

Seasons f Rel f Required angle =

(prop.) Rel f (prop.) × 360o
Winter 72 .72 259.2o

Autumn 15 .15 54o

Spring 10 .10 36o

Summer 3 .03 10.8o

n = 100
§  Both  the  cumula=ve  frequency  distribu=on  and  the  
cumula=ve  percentage  distribu=on  can  be  presented  in  
graphic  form.  The  general  procedure  is  the  same  for  both.  
§  Cumula:ve  percentage  curve  is  graphed  by  plomng  
cumula=ve  percentage  above  the  upper  real  limits  for  each  
class  interval.  (In  cumula=ve  frequency  curve,  we  plot  the  
cumula=ve  frequency,  instead  of  cum%,  on  the  Y-­‐axis).  
§  The  points  are  then  connected  with  straight  lines  and  the  
curve  is  brought  down  to  0  at  the  lower  end  at  the  upper  
real  limit  of  the  preceding  class  interval  (in  which  there  
were  zero  scores).  
§  A  cumula=ve  percentage  curve  never  has  a  nega=ve  slope  
(that  is,  it  never  comes  down).  For  class  intervals  in  which  
there  are  zero  scores,  the  cumula=ve  percentage  curve  
remains  horizontal.    

§  The  curve  shows  a  small  rise  in  intervals  with  rela=vely  few  
scores  and  a  sharp  rise  in  intervals  with  many  scores.  
§  Many  distribu=ons  have  most  of  the  cases  in  the  middle  
por=on  of  the  distribu=on.  This  results  in  a  cumula=ve  
percentage  curve  with  an  S-­‐shaped  figure  called  an  ogive  
curve.  

§  When  the  curve  is  carefully  drawn  and  the  scale  divisions  
are  precisely  marked,  we  can  determine  percen=les  and  
percen=le  ranks  from  the  cum%  curve.  (Connec=ng  the  dots  
on  the  cumula=ve  curve  with  straight  lines  is  the  graphic  
equivalent  of  assuming,  as  we  did  when  compu=ng,  that  
scores  are  evenly  spread  throughout  the  interval).  
Apparent limits Real limits f cum f cum %
96-98 95.5-98.5 1 50 100
93-95 92.5-95.5 0 49 98
90-92 89.5-92.5 2 49 98
87-89 86.5-89.5 7 47 94
84-86 83.5-86.5 10 40 80
81-83 80.5-83.5 6 30 60
78-80 77.5-80.5 8 24 48
75-77 74.5-77.5 4 16 32
72-74 71.5-74.5 3 12 24
69-71 68.5-71.5 4 9 18
66-68 65.5-68.5 3 5 10
63-65 62.5-65.5 0 2 4
60-62 59.5-62.5 2 2 4
N=50
Cumulative percentage curve of the grouped history exam scores
 X f cum f cum %
195-199 1 50 100
190-194 2 49 98
185-189 4 47 94
180-184 5 43 86
175-179 8 38 76
170-174 10 30 60
165-169 6 20 40
160-164 4 14 28
155-159 4 10 20
150-154 2 6 12
145-149 3 4 8
140-144 1 1 2
N=50
§  Grouping  -­‐  There  is  no  such  thing  as  the  graph  of  a  given  
set  of  data.  The  same  set  of  raw  scores  may  be  grouped  
in  different  ways,  and  the  grouping  will  affect  the  graph  
of  the  distribution.    
§  For  example,  the  two  graphs  (a)  and  (b)  show  the  results  
of  the  history  exam  scores,  but  grouped  with  interval  
widths  of  2  and  4,  respectively.    
§  Compare  the  appearance  of  these  two  graphs  with  each  
other  and  with  the  figure,  where  i=3.  
When i =2
When i =4
When i =3
§  Relative  scale  –  The  decision  about  relative  scale  is  
arbitrary,  and  the  resulting  graph  can  be  squat  or  
slender  depending  on  the  choice.  
§  The  large  difference  in  appearance  created  by  
differences  in  scale  is  the  reason  for  the  convention  
that  the  height  of  the  figure  should  be  about  three-­‐
quarters  of  the  width.  
§  Scale  used  for  the  Y-­‐axis  –  Even  if  the  convention  
regarding  height  and  width  is  followed,  the  same  data  
can  appear  very  different  when  graphed,  depending  on  
the  scale  of  measurement  used  for  frequency.  
§  You  will  sometimes  see  graphs  with  a  break  in  the  
vertical  axis.  Never  do  this!  Frequency  on  the  vertical  axis  
should  always  be  continuous  from  zero.  When  we  put  a  
break  in  the  axis,  we  lose  the  proportional  relationship  
among  class  interval  frequencies.  
 Certain  shapes  of  frequency  distributions  occur  with  
enough  regularity  in  statistical  work  that  they  have  names.  
The  names  effectively  summarize  the  general  
characteristics  of  the  distribution.  

§      In  a  rectangular  distribution,  there  

are  equal  number  of  cases  in  all  
class  intervals.      
§  In  skewed  distributions,  one  tail  slants  to  the  right  
(positively  skewed)  or  to  the  left  (negatively  skewed).    

A  negatively  skewed  distribution  results,  for  example,  if  the  participants  

are  given  a  very  easy  test.  Because  most  of  the  participants  score  high  
and  only  a  few  score  low,  the  longer  tail  trails  off  toward  the  X  and  Y  
intercept.  A  positively  skewed  distribution  results  if  the  test  is  very  hard.    
§  In  a  bimodal    distribution,  there  
are  two  peaks  or  humps,  each  
with  the  same  maximum  
frequency.  

§  A  graph  with  three  or  more  humps,  each  with  the  same  
maximum  frequency,  is  multimodal.  
§  Technically,  a  distribution  is  bimodal  or  multimodal  only  if  
its  humps  have  the  same  frequency.  Nevertheless,  
distributions  with  pronounced  but  slightly  unequal  humps  
are  commonly  described  as  bimodal  or  multimodal.    
§  The  figure  below  shows  a  bell-­‐shaped  distribution.    A  
specific  type  of  bell-­‐shaped  distribution,  called  the  
normal  curve,  is  of  great  importance  in  statistical  
inference.  
§  Kurtosis  refers  to  the  degree  of  peakedness  of  a  graphed  
distribution.  
§  The  normal  distribution  is  mesokurtic;  meso-­‐  means  
intermediate.    
§  A  distribution  flatter  than  the  normal  curve  is  called  as  
platykurtic.  It  is  called  leptokurtic  if  it  is  more  peaked  than  
the  normal  curve.