Survey Data Analysis in SPSS: t-Tests, ANOVA, Correlation and Regression

12 pages · 1,879 words

This APA statistics homework for an undergraduate course analyses a survey of 500 potential customers for a company planning to sell tablets in the Phoenix market. It sets out SPSS frequency tables for 13 variables and a reliability check, then tests for differences with two independent-samples t-tests and two one-way ANOVAs, looks for links with bivariate correlations and closes with two linear regressions, explaining after each test what its p value means for the null hypothesis. A table summarizes all eight tests, and charts show party affiliation and mean income by sex.

Introduction

Market segment analysis is one of the critical steps in assessing a new market for a product. Understanding the growth and potential of a market segment requires decisions based on existing data about potential customers’ preferences and purchasing power. With relevant data, a company can assess the internal and external factors that are critical to the success and growth of its business. This report is based on an online survey of potential customers in the Phoenix market, where the company plans to expand its tablet business. It presents the findings of the data analysis and shows how the company can use them to serve its customers.

Four sets of statistical procedures were used to extract useful information from the data set. First, descriptive statistics were generated to understand the structure of the data and to find information to apply in the inferential tests. Second, independent-samples t-tests and one-way ANOVAs compared means to see how different independent variables (IVs) affect different dependent variables (DVs). Third, bivariate correlations tested the relationships among variables that can inform decisions on product lines for Phoenix customers. Finally, linear regression measured the association between a DV and several IVs, to show the business how far each variable predicts a dependent variable of interest.

The findings of this report will help the company base its decisions on a better understanding of its customers in the Phoenix area.

Descriptive Statistics

Sample and Variables

The following variables were selected for this analysis: income, sex, internet hours, petowner, webable, political, homo_marry, religious, facebook, nextgen, companies, banks, science_rsch, USETECH, issues, degree, tvhours, careself, age, emailhr and CINTV.

There were 500 respondents in total. Of the 499 who gave their age, the mean age was 52.27, the median 50 and the mode 42. Ages ranged from 9 to 102; an age of 9 in a survey of adult customers is probably a data-entry error and should be checked before the data are used further. Of the 500 respondents, 274 (54.8%) were female and 226 (45.2%) were male (Table 1). Family income ranged from $383 to $250,000, with a mean of $60,752.71.

Table 1: Respondent’s Sex
ResponseFrequencyPercentValid percentCumulative percent
Female27454.854.854.8
Male22645.245.2100.0
Total500100.0––

Religion, Party and Education

The largest group of respondents (40.0%) were moderately religious, followed by those who were not religious (24.3%). Table 2 shows how religious the respondents considered themselves.

Table 2: Does the Respondent Consider Themselves a Religious Person?
ResponseFrequencyPercentValid percentCumulative percent
Very religious7114.214.314.3
Moderately religious19939.840.054.3
Slightly religious10621.221.375.7
Not religious12124.224.3100.0
Valid total49799.4100.0–
Missing: don’t know20.4––
Missing: no answer10.2––
Total500100.0––

Most respondents who named a party were Democrats (40.9%) or independents (40.5%), as Table 3 and Figure 1 show.

Table 3: Political Party Affiliation
ResponseFrequencyPercentValid percentCumulative percent
Democrat20340.640.940.9
Independent20140.240.581.5
Republican8016.016.197.6
Other party122.42.4100.0
Valid total49699.2100.0–
Missing: no answer40.8––
Total500100.0––

Pick a party to see what it means for the analysis.

Democrat: 40.9%

The largest group, 203 respondents. Party is the grouping variable of the second one-way ANOVA.

Figure 1: Party Affiliation of the 496 Respondents Who Named a PartyValid percentages from Table 3 (four respondents gave no answer). Chart by Writers Per Hour.
Party affiliation of the respondents who named a party, valid percent
PartyShare of respondents (%)What it means for the analysis
Democrat40.9%The largest group, 203 respondents. Party is the grouping variable of the second one-way ANOVA.
Independent40.5%Almost as many as Democrats, 201 respondents, so independents are a market segment of their own rather than a remainder.
Republican16.1%80 respondents, about one in six of those who named a party.
Other party2.4%12 respondents, too few to compare with the three main groups on their own.

Table 4 shows the highest degree of each respondent. Only 118 respondents answered this question; among them, most (55.1%) held a bachelor’s degree.

Table 4: Highest Degree the Respondent Has Earned
ResponseFrequencyPercentValid percentCumulative percent
Associate’s163.213.613.6
Bachelor’s6513.055.168.6
Master’s275.422.991.5
MBA20.41.793.2
Law20.41.794.9
PhD40.83.498.3
MD20.41.7100.0
Valid total11823.6100.0–
Missing: not asked38276.4––
Total500100.0––

Attitudes and Confidence

Most respondents agreed (61.4%) or strongly agreed (27.7%) that scientific research is necessary and should be supported by the federal government (Table 5).

Table 5: Scientific Research Is Necessary and Should Be Supported by the Federal Government
ResponseFrequencyPercentValid percentCumulative percent
Strongly agree6913.827.727.7
Agree15330.661.489.2
Disagree193.87.696.8
Strongly disagree81.63.2100.0
Valid total24949.8100.0–
Missing: not asked23947.8––
Missing: don’t know102.0––
Missing: no answer20.4––
Total500100.0––

Attitudes to self-reliance were mixed: 22.8% agreed that those in need have to take care of themselves, 24.6% were neutral and 52.5% disagreed (Table 6).

Table 6: Those in Need Have to Take Care of Themselves
ResponseFrequencyPercentValid percentCumulative percent
Strongly agree81.62.42.4
Agree6813.620.422.8
Neither agree nor disagree8216.424.647.4
Disagree13126.239.386.8
Strongly disagree448.813.2100.0
Valid total33366.6100.0–
Missing: not asked16432.8––
Missing: don’t know20.4––
Missing: no answer10.2––
Total500100.0––

Most respondents had only some confidence in banks and financial institutions (51.5%); 10.2% had a great deal of confidence and 38.3% hardly any (Table 7).

Table 7: Confidence in Banks and Financial Institutions
ResponseFrequencyPercentValid percentCumulative percent
A great deal357.010.210.2
Only some17635.251.561.7
Hardly any13126.238.3100.0
Valid total34268.4100.0–
Missing: not asked15531.0––
Missing: don’t know30.6––
Total500100.0––

Few respondents had a great deal of confidence in television (16.2%), and almost half (48.1%) had hardly any (Table 8).

Table 8: Confidence in Television
ResponseFrequencyPercentValid percentCumulative percent
A great deal8016.016.216.2
Only some17735.435.851.9
Hardly any23847.648.1100.0
Valid total49599.0100.0–
Missing: don’t know40.8––
Missing: no answer10.2––
Total500100.0––

Media and Technology Use

About a third of the respondents (35.6%) watched television for more than three hours a day, and the mean was 3.32 hours a day. Respondents spent a mean of 14.37 hours a week on the internet. On average they used technology for 51.66% of their time, and 50% was the most common answer. Most respondents rated their ability to use the internet as excellent or good (59.2%); fewer than 100 (17.8%) rated it poor or very poor (Table 9).

Table 9: Respondent’s Ability to Use the Internet
ResponseFrequencyPercentValid percentCumulative percent
Excellent12525.025.025.0
Good17134.234.259.2
Fair11523.023.082.2
Poor5911.811.894.0
Very poor306.06.0100.0
Total500100.0––

Of the 160 respondents asked, 76.9% understood at least some of the issues facing the country, and only 2.5% understood none of them (Table 10).

Table 10: How Well the Respondent Understands the Issues Facing the Country
ResponseFrequencyPercentValid percentCumulative percent
None40.82.52.5
A little336.620.623.1
Some6913.843.166.3
Quite a bit428.426.392.5
A great deal122.47.5100.0
Valid total16032.0100.0–
Missing: not asked33366.6––
Missing: don’t know51.0––
Missing: no answer20.4––
Total500100.0––

Most respondents asked (89.5%) disagreed that science and technology give more opportunities to the next generation (Table 11).

Table 11: Science and Technology Give More Opportunities to the Next Generation
ResponseFrequencyPercentValid percentCumulative percent
Strongly agree51.02.02.0
Agree214.28.510.5
Disagree14428.858.368.8
Strongly disagree7715.431.2100.0
Valid total24749.4100.0–
Missing: not asked24048.0––
Missing: don’t know122.4––
Missing: no answer10.2––
Total500100.0––

More than half of the respondents (51.2%) visited Facebook about once a day, while 8.8% used it every few weeks or less often (Table 12).

Table 12: How Often the Respondent Visits or Uses Facebook
ResponseFrequencyPercentValid percentCumulative percent
Several times a day13426.826.826.8
About once a day25651.251.278.0
A few times a week6613.213.291.2
Every few weeks112.22.293.4
Less often214.24.297.6
Never122.42.4100.0
Total500100.0––

Most respondents (344, or 68.8%) owned pets (Table 13).

Table 13: Is the Respondent Currently a Pet Owner?
ResponseFrequencyPercentValid percentCumulative percent
Yes34468.868.868.8
No15631.231.2100.0
Total500100.0––

Reliability of the Confidence Items

A reliability test (Cronbach’s alpha) was run on three confidence items: confidence in banks, in major companies and in television. The alpha for the three items was .58, below the .70 usually taken as acceptable, so they do not yet form a reliable scale. The item-total statistics show why: confidence in television did not correlate with the other two items (corrected item-total correlation −.03), and deleting it would raise alpha to .91. Confidence in banks and in major companies can therefore be combined into one reliable measure, while confidence in television should be analysed on its own.

Tests of Difference

Independent-Samples t-Test 1: Household Income by Sex

An independent-samples t-test compares the means of two independent groups to determine whether the difference between them is significant. This test compared the mean family income of female and male respondents.

  • Independent variable: sex
  • Dependent variable: income

The null hypothesis was that there is no difference in mean income between male and female respondents. The test included 254 female and 214 male respondents who gave their income. The mean income of female respondents was $55,004.83 (SD = 48,180.02), and that of male respondents $67,574.95 (SD = 51,815.71). Levene’s test was not significant (F = 0.58, p = .448), so equal variances were assumed. The difference of $12,570.11 was significant, t(466) = −2.72, p = .007, and the 95% confidence interval of the difference ($3,476 to $21,664 in favor of men) does not include zero. The null hypothesis was therefore rejected: male respondents reported a significantly higher household income than female respondents.

Editorial addition · not part of the original paperMean Household Income by SexEach bar is the mean family income in constant dollars of one group in t-test 1. The gap of $12,570 is the mean difference the test found significant.

Mean family income in constant dollars, US dollars

  1. Male: 67,575 US dollars.
  2. Female: 55,005 US dollars.

Means rounded to the nearest dollar (SPSS prints $67,574.95 and $55,004.83). Respondents who gave no income are left out, so the two groups add up to 468, not 500.

Chart by Writers Per Hour from the group statistics of the paper’s independent-samples t-test 1 (SPSS output).

Show every figure as a table
Mean family income by sex, t-test 1
SexMean income (US$)RespondentsStandard deviation (US$)
Male67,57521451,815.71
Female55,00525448,180.02

Independent-Samples t-Test 2: Internet Hours by Pet Ownership

This t-test compared the hours a week spent on the internet by respondents who owned pets and by those who did not. The null hypothesis was that the means of the two groups are equal.

  • Independent variable: petowner
  • Dependent variable: internet_hrs

Pet owners spent a mean of 14.51 hours a week on the internet and respondents without pets 14.06 hours, a difference of 0.45 hours. Levene’s test was not significant (F = 0.05, p = .829), so equal variances were assumed. The difference was not significant, t(498) = 0.47, p = .637, so the null hypothesis was retained: owning a pet does not affect the hours a person spends on the internet each week.

One-Way ANOVA

ANOVA 1: Email Hours, Work Hours and Technology Use by Internet Ability

A one-way ANOVA compares the means of a dependent variable across the groups of an independent variable. This ANOVA tested whether email hours a week, work hours a week and the percentage of time spent using technology differ by the respondent’s ability to use the internet.

  • Independent variable: webable
  • Dependent variables: emailhr, HRS2 and USETECH

None of the three differences was significant: email hours, F(3, 370) = 0.18, p = .910; work hours, F(3, 371) = 1.55, p = .201; and technology use, F(3, 364) = 1.36, p = .254. A respondent’s ability to use the internet is therefore not linked to the hours they spend on email or at work, or to the share of their time they spend using technology.

ANOVA 2: Same-Sex Marriage and Self-Reliance by Party

  • Independent variable: political
  • Dependent variables: homo_marry and care_self

Both differences were significant. Agreement that those in need have to take care of themselves differed by party affiliation, F(3, 325) = 4.94, p = .002, and so did agreement that homosexuals should have the right to marry, F(3, 326) = 11.09, p < .001. Respondents’ views on self-reliance and on same-sex marriage therefore depend on their political affiliation, with a larger effect for same-sex marriage.

Bivariate Correlations

Correlation 1: Religiosity, Facebook Use and Technology Optimism

Religiosity was not significantly correlated with how often respondents used Facebook, r(495) = −.05, p = .226, or with the belief that science and technology give more opportunities to the next generation, r(251) = .10, p = .115. Facebook use and that belief were weakly but significantly correlated, r(252) = .13, p = .039. Both items are coded from the most frequent use and the strongest agreement upward, so respondents who used Facebook less often were slightly more likely to disagree that technology opens opportunities for the next generation. Facebook use, not religion, is thus linked, if weakly, to optimism about technology.

Correlation 2: Confidence in Companies and Banks and Support for Science Funding

Confidence in major companies and confidence in banks and financial institutions were moderately and significantly correlated, r(335) = .36, p < .001: respondents who trusted one tended to trust the other. Confidence in major companies was not correlated with support for federal funding of scientific research, r(242) = .03, p = .670. Confidence in banks had a weak but significant negative correlation with that support, r(245) = −.15, p = .021; given the coding, respondents with less confidence in banks were slightly more likely to agree that the federal government should fund research.

Linear Regression

Regression 1: Technology Use on Income, Understanding of Issues and Degree

  • Independent variables: income, issues and degree
  • Dependent variable: USETECH, the only continuous dependent variable in the data set

Only 44 respondents had answers to all four variables, because degree and understanding of issues were asked of few respondents. The model was not significant, R² = .088, adjusted R² = .020, F(3, 40) = 1.29, p = .291, and none of the three predictors was significant on its own (all p > .05). Income, understanding of current issues and level of education together explain less than 9% of the variation in the share of time respondents spend using technology, and the model is no better than chance.

Regression 2: Technology Use on TV Hours, Confidence in TV, Self-Reliance and Age

  • Independent variables: tvhours, CINTV, careself and age
  • Dependent variable: USETECH

This model was not significant either, R² = .012, adjusted R² = .000, F(4, 316) = 0.98, p = .418, and none of the four predictors was significant (all p > .05). Hours of television, confidence in television, attitude to self-reliance and age do not predict how much of their time respondents spend using technology.

Conclusion

The analysis of 500 survey responses gives the company a mixed picture. Four of the eight inferential tests found significant results. Male respondents reported a significantly higher household income than female respondents. Attitudes to self-reliance and to same-sex marriage differed by political affiliation. Confidence in major companies went together with confidence in banks, and two weak correlations linked less frequent Facebook use with doubts about the opportunities technology brings, and lower confidence in banks with greater support for federal science funding. The other tests found no significant differences or links: internet hours did not differ by pet ownership; email hours, work hours and technology use did not differ by internet ability; and neither regression model predicted the share of time respondents spend using technology. Table 14 summarizes the results.

Table 14: Summary of the Inferential Tests
TestVariablesResultSignificant at .05?
t-test 1Income by sext(466) = −2.72, p = .007Yes
t-test 2Internet hours by pet ownershipt(498) = 0.47, p = .637No
ANOVA 1Email hours, work hours and technology use by internet abilityF = 0.18, 1.55 and 1.36; p = .910, .201 and .254No
ANOVA 2Self-reliance and same-sex marriage by partyF(3, 325) = 4.94, p = .002; F(3, 326) = 11.09, p < .001Yes
Correlation 1Religiosity, Facebook use, technology optimismOnly Facebook use and technology optimism: r = .13, p = .039One of three
Correlation 2Confidence in companies and banks, support for science fundingCompanies and banks: r = .36, p < .001; banks and science: r = −.15, p = .021Two of three
Regression 1Technology use on income, understanding of issues, degreeR² = .088, F(3, 40) = 1.29, p = .291No
Regression 2Technology use on TV hours, confidence in TV, self-reliance, ageR² = .012, F(4, 316) = 0.98, p = .418No

For the tablet launch in Phoenix, household income, which differs by sex, matters more for pricing and promotion than pet ownership or political views, and frequent Facebook users are a slightly more receptive audience for messages about technology. Because none of the measured variables predicted technology use, the company should collect direct data on tablet use and purchase intentions before building its product lines.

Rate This Sample

How useful was this example?

Your rating

Cite This Sample

Survey Data Analysis in SPSS: t-Tests, ANOVA, Correlation and Regression. (n.d.). Writers Per Hour. https://writersperhour.com/samples/homework/survey-data-analysis-in-spss

We'll ensure your essay is what you asked for

  • Have skilled academic writers complete your assignment.
  • Pay a fair rate for a writing that follows your guidelines.
  • Obtain a 100% unique essay ready before your due date.