Research Design Lecture Notes

Analyzing raw data:

Data collection in excel

-pt number in 1st column and then characteristics along horizontal...

-convert everything to numbers

Variable types

-categorical

-binomial

-ordinal

-nominal eg race

-continuous (non-ordinal- the difference bn each step is the same...)

Parametric -normally distrib. Ok to use t test, mean, etc

Nonparametric- everything else (race, sex, and maybe age if not normally distributed...., length of stay is rarely parametric...)

Data Types

Independent

Paired- eg per and post Glenn...

Clustered, Longitudinal-

Centrality

-Mean

-Median- best if things are skewed or has outliers that might change the mean...

Variability

Standard deviation- only good if data is normally distrib

Alternates:

-range- good if small n

-inter quartile range- show the 25th & 75th %- gives a similar idea to standard deviation but for non normally distrib data

Analytic Stats

What is the predictor and what is the outcome?

Bi x bi -- x2 or fisher exact. Fisher is better but x2 gives better t test, if you have under 5 items in each box u must do fisher exact

Non parametric tests

Chi sq

Chi sq trend -for ordinal variables

Mann Whitney aka wilcoxan rank -one nonpar var bn 2 groups, instead of t test

Kruskal wallis. If compare nonpara over multiple groups, more than 2

Spearman rank -correlation coefficient

McNamara test - paired binomial.

Parametric tests

Independent t test- the two var are independent

Paired t test- eg comparing subject to self

A nova - like krskll wallis...

Pearson coefficient

--see table fr the ppt slide for summery