QM 2 ASSUMPTIONS AND WHAT TESTS

one population test?

Parametric: one sample t test

Non-parametric: sign test or wilcoxon signed rank test

Paired/dependent samples

Parametric: paired sample t-test

Non-parametric: sign test or wilcoxon signed rank test

Two independent populations equal variances

Parametric: two sample pooled t-test

Non-parametric: wilcoxon signed rank test

Two independent populations unequal variances

Parametric: welch two sample t-test

Non-parametric: wilcoxon signed rank sum test

Exact proportion one population test statistic assumptions

• data is a random sample of independent observations

• variable of interest is qualitative and has only 2 mutually exclusive but exhaustive categories

• the measurement scale is nominal or ordinal

• sampled population is binomially distributed

Exact proportion 2 population proportion assumptions

• data is a random sample two independent randome samples of observations

• variable of interest is qualitative and has only 2 mutually exclusive but exhaustive categories

• the measurement scale is nominal or ordinal

• sampled populations are binomially distributed

Assumptions for one sample t test

• data is a random sample and samples are independent

• voi is quantitative and continuous

• is measured on a ratio or interval scale

• sampled population is approximately normally distributed

Assumptions for sign test

• data is a random sample of independent observations

• VOI is quantitative or qualitative

• measurement scale is at least ordinal

Assumption for wilcoxon signed rank test

• data is a random sample of independent observations

• VOI is quantitative and continuous

• measurement scale is interval or ratio

• distribution of sampled pop is symmetric

Assumptions for matched pair sampling parametric

• data is a random sample and samples are independent pairs of observations

• voi is quantitative and continuous

• is measured on a ratio or interval scale

• sampled population of differences is approximately normally distributed

Assumptions for matched pair sampling non-parametric (wilcox)

• data is a random sample of independent observations

• VOI is quantitative and continuous

• measurement scale is interval or ratio

• distribution of sampled pop differences is symmetric

Assumptions for independent samples parametric

• data is a random sample and there are 2 independent random samples

• voi is quantitative and continuous

• is measured on a ratio or interval scale

• sampled population of differences is approximately normally distributed even though variances are unknown

Assumptions for indepenedent samples non-parametric (wilcox)

linear regression assumptions

1. we have a random sample of n>k+1 independent obs that satisfy pop regression model

2. each random error has 0 conditional expected value, ie error terms are not correlated with IVs

3. conditional variance of error term is constant implying overall var is constant

4. conditional var of any two random errors is 0

5. no linear relationships between rhs variables

6. conditional distrubution of random errors is normal