Week 5 Common Stats Testing PDF

Preview

Summary

🛷Week 5 : Common StatsTesting5 Choosing the right statistical testWhat is the main research question? What type of data will help answer the research question? Scale of measurement Dependent vs. Independent Number of variables/groups to be compared. Dependency of groups5 Research question and design...

Description

Week 5 : Common Stats Testing 5.1 Choosing the right statistical test What is the main research question? What type of data will help answer the research question? Scale of measurement Dependent vs. Independent Number of variables/groups to be compared. Dependency of groups 5.2 Research question and design e.g. Descriptive studies - designed to explore prevalence - don't need complex stats tests Analytical studies

Paired sample t-test : groups that are dependent on each other , before and after studies, u have a grp of ppl and u wanna see if there is a change before and after the intervention, so u look within the grp for changes Correlation and regression analysis : association/ relationship btw bp and how much exercise u do Survival analyses : Time to event , how long it takes btw when a person gets diagnosed till their death or when they get treatment 5.3 Type of variable Dependent(outcome) vs Independent (exposure) No of individuals in each group - some studies only cater to a certain no, any more than that may cause lots of bias and errors in analysis Operational definition of each variable - things are defined differently across diff research papers For Obesity weight vs using BMI Level of measurement : Ordinal, Nominal, Interval, Ratio Normal distribution? Bell curved? 5.4 Type of test Parametric tests make a number of assumptions about the distribution of the population from which the sample is drawn (e.g. normally distributed scores) and the nature of the data(interval or ratio level) e.g. Paired sample t-test, Independent T-test, Annova, Pearsons's correlation Non-parametric tests do not make assumptions about the distribution of the data - more suitable techniques for smaller samples or for when the data

collected is measured on the ordinal or nominal level. e.g. Chi-square test , Mann-Whitney U test, Wilcoxon signed rank test, Kruskal - Wallis, Friedman, Spearman's correlation Weakness of non-parametric — Less powerful than parametric - less likely yo detect true differences/ relationships that exist Difficult with interpretation : use ranks rather than raw value 5.5 Correlation measure of strength and direction of the association between two variables Pearson's correlation : strength btw 2 interval/ratio level variables Spearman rank correlation Spearman's rho) : compares 2 ordinal level variables Point biserial : correlation btw categorical and continuous data Correlation coefficient r 0 = No relationship 1 = Perfect negative correlation 1 = Perfect positive correlation

Negative

Interpreting correlation

Positive

No correlation

r = 1 is a very strong correlation and since it is (+) , it is a positive correlation → Strong perfect positive correlation

r=0 , no correlation

r= 0.90 → Negative correlation r=0.5 Moderate correlation

Correlation does not mean causation

5.6 Overview

If your measurement is of MEANS, Are u gonna compare the means , are u interested in averages, in collecting continuous data of a grp of ppl and u wanna compare their averages 1 group followed before and after: Matched) Paired TTest 2 groups Intervention vs Control): Two (independent) Sample T-test 2 Groups: ANOVA (variance)

If your measurement is of Proportion, Frequency, are u just counting the no. Of events, is the outcome data nominal or categorical data, or raw count of event TABLE Large Group): Chi Squared Small 10 Group: Fisher’s Exact Test

5.7 TTests Independent sample t-test/2 sample t-test When we have two samples that are independent of each other and we want to compare the means of the two samples. For example, you want to compare cholesterol levels between runners and swimmers so u look at HDL levels in both and compare the average of means compare means of 2 independent groups Dependent variable: Continuous Independent variable: Categorical

*IF p value 0.05, there is sig. diff btw means the 2 groups. Interpretation : Report the means of the two groups or the mean difference and confidence interval & p value Paired sample t-test When we have two samples that are NOT independent of each other and we want to compare the means. For example, you want to know if a certain diet is effective at reducing LDL cholesterol levels (the bad kind of cholesterol). Could apply to pre- and postdata. E.g. mean cholesterol levels pre- and post- starting a diet plan (each individual has a pair of data , pre and post) DV : Continuous time point Mean diff btw 2 time points and P values Used only when data is paired/ matched To compare before/after measurements of the same variable To compare how a group of subjects perform under two different test conditions. *IF p value 0.05, there is sig. diff btw 2 time points/experiments. Interpretation : Look at p value and mean difference & CI values

Reporting of t-tests

5.8 Chi Square tests (x2) outcome data are nominal only Assesses whether two variables are independent Any number of groups 2 X 2, 2 x 3, 3 x 3 , etc) For some researchers, Anything (sample size) less than 2030 is small Anything more is large At least 40 ppl should be there to use chi sq test (it varies for researchers) Comparing proportions or raw numbers E.g. Comparing heart attack events by gender Comparing of pass and fail rates on three testing sites Comparing the percentage of men and women across different age groups(< 18, 1829, 3039, 4049, 5060, > 60 admitted to the hospital for covid-19 treatment Compare % , proportions, frequencies across 2 or more groups DV Categorical IV : Categorical *If p < 0.05, there is significant evidence that the proportions are not the same. Interpretation : Use %’s to describe what the relationship is. Look at % and p value.

Examples : Is there a r/s btw heart disease and gender? What do u have : One categorical IV - gender, One categorical DV - diagnosis of heart disease (yes,no) Another example : most appropriate investigate if social class (deprived, middle class, affluent) differed between patients attending two different hospitals? How is Chi square reported?...