Question

1) Look at the statistical analyzes that I have provided. Now, identify the next necessary step related to the statistical conclusions I have reached. (Hint: It has to do with t-tests)

Here are the results of the Observation Study data analysis.

Descriptive Statistics

Independent Variable 1 - Gender: a categorical variable

N = 1042 (frequencies: Male = 469, Female = 573)

Independent Variable 2 - Behavior: a categorical variable

N = 1042 (frequencies: Cell = 184, MP3 = 84, None = 774)

Dependent Variable - Looks: a continuous measure

# Looks	0	1	2	3	4	5	6	7
Frequency	186	306	346	123	52	21	5	3

N = 1117 Mean = 1.66 Standard Deviation = 1.242 Range = 7

Inferential Statistics:

I ran a 2 (Gender: Male vs. Female) X 3 (Behavior: Cell vs. MP3 vs. None) ANOVA with the number of looks as the Dependent Variable. Here are the results…

Source	Sum of Squares	df	Mean Square	F	Sig.
Gender	2.157	1	2.157	1.435	.231
Behavior	35.298	2	17.649	11.745	.000
Gender * Behavior	11.951	2	5.975	3.976	.019
Residual	1556.788	1036	1.503
Total	4481.000	1042

Condition Means (marginal means in bold)

	Cell	MP3	None
Men	1.714	2.000	1.694	1.73
Women	1.800	2.528	1.485	1.61
	1.76	2.23	1.58

Answer 1

From the results of the two-way ANOVA, we can conclude that there is no significant differences in the Number of looks between males & females as the p-value=0.231>0.05.

The behaviour variable is significant as the p-value = 0.000<0.05 is significant for it. Therefore, it means that atleast two groups within Behaviour have significant differences in the number of looks but we are not aware of which groups. Hence, for this purpose, we need to run a post-hoc test. As we have three levels within Behaviour which are (Cell, MP3, None), LSD post-hoc test would be appropriate here. The post=hoc results would tell us that which levels have a significant difference between each other. Also, we could run a post-hoc test on the interaction variable to know which level of the factors interact with each other.

The post-hoc tests check the difference between two groups pairwise so it is related to t-tests but adjusts the familywise error rate to 0.05 rather than individually adding up the error rates for each pairwise comparison done.