Path: janda.org/c10 > Advice > Formulating a Hypothesis
```

Statistics Research Paper
Formulating a Hypothesis

Steps in Formulating a
Hypothesis

Decide what you want to
explain: choose a dependent variable

variable must show variation

Run
Descriptives to see mean and dispersion
statistics

Even better, run
Frequencies, and call for a histogram
along with the mean and std. dev (suppress the
Frequency table itself)

Choose
independent variables that also show
variation

One can't explain
variation in a dependent variable with an independent
variable that doesn't vary.

Moreover, the
variation in the independent must match that in
the dependent variable.

Otherwise, they can't
possibly covary, which is needed for the
covariation needed in correlation.

Examples
of lack of theoretical potential:

Voting for Clinton
in 1996 in the suburb of Winnetka can't be
explained by race, for race does not vary much in
Winnetka.

Variation across
time in deaths in domestic violence within
nations can't be explained by ethnic diversity, for
ethnicity doesn't change much across time within a
nation.

However, variation
in deaths across nations can be
related to variation in ethnic diversity across
nations.

Think of multiple
causes of the dependent variable:

Do two or more
independent variables combine to affect it?

Consider using
multiple regression to deal with
multiple causes.

Does a relationship
hold for some units of analysis but not others?

For northern
states but not for southern ones?

For European
nations but not for Third World
nations?

For whites, but
not for blacks?

For reformed
cities, but not unreformed cities?

For multi-party
states, but not single-party states?

Try to develop your
analysis so that it considers all the cases,
even if the relationship doesn't apply equally to
them.

Consider alternative
measures of both the dependent and independent
variables.

```