Environmental Ethics Survey Data Analysis Exercise
1. First look at the questionnaire and try and remember what we were thinking in putting this together! What are the groups of questions, what other kinds of information did we gather, what kinds of questions might you have about how people responded.
2. Download the excel file from D2L onto your computer and save it in your H drive file. It’s good to save it twice (two labels)… one as an archival file and one that you will manipulate and where you can make mistakes without permanent damage.
3. Look over your working copy… (do this again later) are there entries that are missing a lot of values? Either get rid of them or segregate them. How could someone get a total score of 40? Are there differences in how people have coded things that will create problems? (shopping preference, occupation, …)? You may do some clean-up, but things do not have to be identical.
4. Now turn to your excel page and make sure you have the Data Analysis ToolPak (see the second page of the handout to do this).
5. Everyone will be required to do a descriptive analysis and a histogram of the total scores (all of the 32 judgments, summed). We’ve created a column of summed responses at the end of the survey (at the far right). Follow these and then discuss what you see in the analysis. I’d like to see a reasonable account of the shape of the distribution, it’s central tendency (Do mean and median differ? What was the most frequent response total?) and variability indicators and where your personal score falls within the group (how many standard deviations above or below the mean are you?).
6. In addition you will carry out an analysis or comparison on some subset of the data of your choosing. Do the rural responses differ those from suburbia? What about political identification? To make these kinds of comparisons, first sort and then do the descriptive analysis on the two (or three subsamples). You may want to see how the responses differ for the responses to a particular question or for a subgroup of questions (animal welfare?). You might also consider a correlation between responses to two questions or two subgroups (is there one question that is a good predictor of the total score?). You may want to copy and paste a column or two into a second sheet.
This is what I want you to do :
7. When you have carried out your analysis, write a reflection on what you see in your results. What does your analysis of the data tell you about people’s “environmental ethic”? Your analysis should include a description of how the survey was constructed and how people were sampled, explaining that this is not intended to be a statistically representative sample. Paste the histograms and the tables of descriptive stats and hypothesis tests into your report. Include any thoughts you have about how the wording, the questions and the sampling might be improved and what questions our findings raise in the realm of environmental ethics.
Sample Analysis for Environmental Ethics Survey Data
I ran descriptive statistics for the very first prompt “Training a dog with a electric shock collar.” Remember the scale was from 1-5 with a 3 defined as moderately objectionable. The mean here was 2.9 (no reason to run it out to six decimal places), median 3 and mode (most common answer) 2. If you haven’t had statistics, probably best to ignore the standard error, standard deviation, and variance- these are all measures of how spread out the answers are. For a large sample we expect about 2/3 of a population to fall within one standard deviation on either side of the mean. Kurtosis and skew tell us something about the shape of the histogram. A nice symmetrical, bell shaped, normal curve would have a score of zero for each. Ours is sort of dome shaped with no good tails—negative kurtosis (extreme statistical trivia). It also looks as if it’s been shoved to one side—values extend out to positive side, positive skew, skewed to the right. Max-Min gives range. When you create a histogram, you’ll want to make a bin table going from min to max at nice even intervals. Here it’s a no-brainer: 1,2,3,4,5. The count is just a check to see you haven’t screwed up. It’s 133 not 134 here because I had one person who left gender blank. They were left out of this analysis.
Q3 Dog Collar Data
Standard Error 0.105588
Standard Deviation 1.217702
Sample Variance 1.482798
You can change the size and proportions of the histogram by clicking on a corner and dragging. Clicking on the legend allows you to change them. Change appears in the function box until you hit enter. In your analysis you should comment on the shape of your histogram.
I sorted the data by gender and then ran a t-test to determine whether there was a significant (p< 0.05) difference between female and male responses. If my hypothesis was specific (females will be more disapproving), I do a one-tailed test. If I don’t specify which way I think it will go ahead of time, it’s a two-tailed test. My Null Hypothesis is zero difference, the means are different (3.15 vs. 2.61) and both the one and two tailed tests are significant (P = 0.0047, P=0.0094 both < 0.05). If your P values are less than 0.01, you get to say the results are “highly significant”. Our results are highly significant… the probability that we would get this big a difference between female and male responses due to chance alone is less than one in a hundred. If you were a dog, how would this affect your kennel behavior?
t-Test: Two-Sample Assuming Equal Variances
responses by gender female male
Mean 3.152778 2.606557
Variance 1.624218 1.175956
Observations 72 61
Pooled Variance 1.418907
Hypothesized Mean Difference 0
t Stat 2.635095
P(T<=t) one-tail 0.004713
t Critical one-tail 1.656569
P(T<=t) two-tail 0.009426
t Critical two-tail 1.978239
When you write it up you can include the tables or not, but you should refer to, report, discuss the key values. How big your sample was, what the means and medians were (skewness will make them diverge), something about the shape of the curve, the presence of any outliers. If you’re comparing two groups tell us what you expected to happen (one-tail or two), what your hypothesis is, give the respective means, and the results of any statistical test you might run. You need not include all the numbers on the chart, but give the means, the t value or other test statistic, the size of the groups, the P value and whether or not any difference is significant. The wording for a significant result is “The probability that the observed difference would occur by chance alone is 0.0094, therefore we reject the null hypothesis that males and females are the same in their responses to this scenario.” If the results were not significant… the probability that the observed difference could occur by chance alone was 0.061 which is not less than 0.05, therefore we cannot conclude that the difference was not due to chance alone. Males and females may not differ in their response to this prompt.