HSCI 190 module 2

bar graphs

summarize categorical data, each bar represents the count and the category

histograms

depict scale data

histograms x axis, y axis

x axis: independent variable is broken into intervals on continuous scale, bars next to eachother

y axis: always start at 0, shows the count of values w/i each interval

width of bar, histogram

represents the size of the interval

area of all bars in a histogram

represents true frequency, because area = 100% of data, absolute and relative freq doesnt matter, it will have the same shape.

unimodial

graph having 1 peak(mode)

bimodial

graph having 2 clear peaks(modes)

multimodial

graph having 3+ peaks(modes)

symmetric graph

mean, median, mode all same at peak

positive skewed graph

mode, median, mean(in that order)

negative skewed graph

mean, median, mode (in that order

frequency polygons

visualizing scale data, same axises as histogram (but x axis marks midpoint of each interval rather than the borders of interval), then a line going thru middle of each bar

- good for comparing multiple groups

cumulative frequency polygon

- the sum of counts up to a certain interval

one way scatter plots

- categorical or scale, uses a single axis to display the relative potition of each point in a group

- advantage: all obs are represented individually

- downside: can becomes hard to read if a lot of points are close together

- can be vertical or horizontal

boxplots

categorical or scale data

- anything beyond adjacent values are considered extreme values and are plotted as individual dots

one way scatter or boxplot for single axis data?

scatter: good to show many obs, maybe overwhelming

box: nice summary of iqr, range, easy to interpret, doesnt tell u obs, excludes extreme values, used more often

two way scatter plots

depict relationship between 2 scale variables

each point represents where the x and y axis meet

- multiple can be used to compare groups (need different colours and a legend)

line graphs

represent relationship between 2 scale variables

- each point on the x axis has a corresponding y value

- most commonly the scale of x axis represents time

what constitutes a potential outlier - mean and stdv

values more than 2 stdvs above or below the mean

what constitutes a potential outlier - median and iqr

values more than 1.5 times the iqr above or below the quartiles

which outlier equation to use?

- median and IQR=most stable bc outliers can impact the mean

- mean and stdv=easier to interpret

data entry errors

types, can cause an outlier, double check data, correct error, re analyze

process error

issue when collecting data, if possible re do data collection, if not possible, remove outlier

when to keep outliers

when they appear to be genuinely obtained and might give new insight on a phenomenon or signal another group that should be accounted for

missing factor

reflect on if there is another factor that might impact outcome that should be considered - reflect on if you should remove or not and explain

random chance

the value is theoretically possible but highly unlikely - reflect on if you should remove or not and explain

random sampling

random selection is used to choose observations

each thing has equal chance of being included in sample

non random sampling

the items included in study are selected for a reason (proximity, feasibilty)

may not give you representative sample of population

sampling bias

when each memeber of relevant population does not have equal chance of ending up in sample

response bias

when participants give answered they believe the researcher wants or socially accepted answers

survirorship bias

when inidviduals leave the study and the researcher continues to measure the remaining participants without considering those that left

recall bias

when participants don’t remember past events properly or omit details, especially when measuring data long time after the event

representative samples

when your data represents the entire population

random sampling benefits and downfalls

can mitigate sampling bias, can ensure representative samples, but challening

transparency

sometimes not possible to use random sampling so you have to be transparent with methods so others can find possible mistakes

central tendency for skewed data

median because it takes extreme values into account but not greatly impacted by them